v0.8.0.3 — Skip on the left, a caption that says who, and a clock that stops

stretching

Three things from playing v0.8.0.2, all about the row rather than the mechanism.

Skip was on the far right and a player's eye is on the countdown. Moved to the
left, in front of the count.

The caption said what but never who. Measured over 40 turns of a real 3-seat game,
half the waiting is automatic phases — 21.0s of phases against 21.7s of other
players — and a phase narrates as "Mainline", which is accurate and no answer at
all to "who am I waiting on". A phase introduces itself now: "The Division:
Mainline phase". A player's move already carries its name from record(), so it is
left alone. The row was also hiding one step early, because it showed only while
behind > 0 — which goes false exactly when the last step of a burst goes up, so the
step most likely to be read lost its caption.

And the speed control was stretching the clock along with the players. It was not
his own move being replayed — own moves have cost nothing since v0.8.0.1 — it was
the phases behind it, which put 105 seconds of clock-ticking into a 5x game. pace
now scales a player's move and leaves a phase at its tabled beat, which is what the
control has always claimed to do. Off still means off for both.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X6cF1iYvJ1kNmzYBzu4QX6

CHANGELOG.md

@@ -19,6 +19,175 @@ page as `v0.1.0 · <sha> · <date>`, so what is deployed can always be identifie
 
 ---
 
+## 0.8.0.3 — 2026-09-09
+
+Three things from playing v0.8.0.2, all of them about the row rather than the mechanism.
+
+### Skip was at the wrong end of the row
+
+Jesse: *"the skip button should be on the far left, in front of where it says [the count], so it's
+always close to where people are looking."* It was on the far right, and a player's eye is on the
+countdown. Moved.
+
+### The caption said what, but never who
+
+*"I saw 2 behind, 1 behind, and then it was caught up, but it didn't tell me what the actual action
+was, like who I was waiting on or what they were doing. I knew I was behind, but I wasn't sure what I
+was supposed to be looking for."*
+
+The caption was there. It was the wrong half of the sentence. **Measured over 40 turns of a real
+3-seat game, half the waiting is automatic phases** — 21.0s of phases against 21.7s of other players —
+and a phase narrates as "Mainline", which is accurate and no answer at all to "who am I waiting on".
+A phase now introduces itself: **"The Division: ▸ Mainline phase"**. A player's move already carries
+its name from `record()`, so it is left alone rather than stuttering it twice.
+
+**And the row was hiding a step early.** It was shown only while `behind > 0` — which goes false the
+moment the LAST step of a burst goes up, so the one step a player was most likely to be reading about
+lost its caption. It now stays up while the queue is still showing something, and reads "catching up"
+once nothing is queued behind.
+
+### The speed control was stretching the clock, not just the other players
+
+*"After my turn, when I actually execute my turn, I'm still subject to that same delay before it
+moves on. That makes no sense. Since I've just done my turn, I don't need to wait after it."*
+
+He was right, and it was not his move being replayed — own moves have cost nothing since v0.8.0.1. It
+was the automatic phases behind it, which were scaling with `pace` along with everything else. At 5×
+that put **105 seconds of clock-ticking** into the game, all of it after a player's own move and none
+of it anything to watch.
+
+**`pace` now scales a player's move and leaves a phase at its tabled beat.** The control is labelled
+as how long another player's move is held, and that is now what it does. A phase still gets its beat
+(TODO #18) and still vanishes entirely at `pace = 0`, because off has to mean off.
+
+---
+
+## 0.8.0.2 — 2026-09-09
+
+Two things found by playing v0.8.0.1 on `phoenix.local`, neither of them in the mechanism itself.
+
+### `?pace=` never worked, and a whole game was played at the wrong speed
+
+Jesse: *"I'm playing at pace = 7, and the bots are still moving too fast for me to follow."* At 7×
+a switching move holds for seven seconds, so that could not be calibration — and it was not. **He was
+at 1× the entire time.**
+
+`index.html`'s two doors are `./play.html?lobby` and `./play.html?solitaire`. Arriving through the
+splash therefore **replaces** the query string, and `location.search` on the play page is `?lobby` —
+so `PACE_OVERRIDE` was null and it fell back to the stored setting of 1. v0.8.0 shipped `?pace=` as
+the only way to change speed and the game's own front door destroyed it. Verified rather than
+assumed: the queue at pace 7 holds a bot's turn for 32.9s with the bot's district up for 24.5s, so
+the mechanism was right and the value never arrived.
+
+Fixed twice over, because one of them is the durable answer:
+
+- **A speed control on the play screen**, beside zoom — `− 1× +`, persisted per viewer, reading
+  through to the queue on the very next move. `PACE_LEVELS` is `0, 0.5, 1, 2, 3, 5, 7, 10`: off is
+  the first rung (TODO #18's "a player who has seen it a hundred times will want it off") and the
+  ladder reaches the speeds people actually reach for. At the top, a six-move switching turn takes a
+  full minute to watch.
+- **The doors now carry `pace` through**, so the URL lever is honest for handing two playtesters
+  different speeds — the only thing it was ever for. When one is present the control says
+  `7× (URL)` and disables itself rather than showing buttons that do nothing.
+
+`PACE_LEVELS` lives in `sim/pacing.ts` with `DWELL` and `MAX_PACE`, not in `main.ts` — the whole
+tuning surface in one file, and testable, which a constant inside the page entry point is not.
+
+**The committed default is unchanged at 1×.** What it should be is a question for a game played at a
+speed that actually took effect.
+
+### "0 today, 2 in all" — the Day-end dialog contradicted itself Jesse, at the end of a Day 1 with
+two collisions in it: *"It shows a total of two collisions, but zero today. Since we just finished day
+one, that does seem to be a contradiction."* Unrelated to v0.8.0 — this has been wrong since the
+dialog was built for Gitea#10, and nobody had played a Day with a collision in it and then read the
+summary.
+
+#### One line of ordering
+
+`advance.ts`, at the rollover:
+
+```ts
+s.clock.day += 1;
+s.collisionsToday = 0;
+```
+
+And `noteDayEnd()` fires when `f.day` goes UP — so the dialog reporting the Day that just finished is
+drawn from the very frame in which that Day's count was zeroed. It printed the *new* Day's zero beside
+a running total that could not possibly agree with it. Reproduced on four of five seeds before
+touching anything: Day 1 ended with `today=3 total=3`, and the dialog read `today=0 total=3`.
+
+**Not derivable on the client, which is why the fix is in the engine.** A Day turns over inside the
+phases that run themselves, so in multiplayer the push announcing the new Day is the same push that
+carries the reset — a client may never see the ended Day's final count to remember it. So
+`collisionsPrevDay` is captured in state at the rollover, immediately before the reset, and rides on
+the frame like the other two counts.
+
+#### And "today" was the wrong word anyway
+
+Even with the right number, a dialog headed "Day 1 has ended" should not say "today" — by then
+"today" is Day 2. It now names the Day: **"Collisions: 2 on Day 1, 2 in all."** The end-of-game
+results screen passes no Day and keeps "today", where the Day has not turned over and the word is
+accurate.
+
+`test/redaction.test.ts`'s allow-list did its job on the way through: adding a public property failed
+the suite until it was declared out loud.
+
+---
+
+## 0.8.0.1 — 2026-09-09
+
+**Bot play was way too fast.** v0.8.0 was installed on `phoenix.local` and played within the hour;
+Jesse: *"I briefly saw that it was the bot's office area then their turn was done and it pointed back
+to my office area"*, and the countdown row appeared "very briefly". Everything else looked right —
+the bots were visibly doing things — so this is calibration and one real bug, not a redesign.
+
+### The bug: the last step of a burst never got its moment
+
+`busy()` was `pending.length > 0`. So the instant the FINAL step of a burst was shown, the queue
+reported itself idle — the animation loop stopped and, because the district panel follows `busy()`,
+it snapped back to the viewer's own board without that step ever being looked at. The countdown row
+went with it. `busy()` is now `pending.length > 0 || dueAt !== null`: there is more to come, **or**
+what is on screen has not had its moment yet.
+
+### The calibration: 250ms was invented, and it was wrong
+
+Jesse's instruction had been "start at 1s and tune down". That was applied to switching and then a
+250ms `action` tier was made up beside it, which held for the case the design was measured against —
+a switching burst — and failed the common one. **Switching is not legal until there is track down**,
+so an early-game bot turn contains none of it. Measured from a real 3-seat game, one bot turn was:
+
+```
+localOps.choose 0ms · draw.fromHomeOffice 250ms · card.play 250ms
+draw.end 0ms · localOps.choose 0ms · freightAgent.stockOutbound 250ms
+```
+
+**750ms for a whole turn.** `action` is now 700ms, which puts that same turn at 4.7s.
+
+**And `localOps.choose` was the worst of it.** It was classed as bookkeeping, at zero — but it is the
+line reading *"Player Bot 1 chose to SWITCH — six Moves to shunt cars around the yard"*: the heading
+for everything that follows. A bot's turn began with no indication of what it was about to do. It is
+an announcement, and it is now in `action`.
+
+### The viewer's own moves cost nothing
+
+Raising `action` exposed a waste: your own click was being held for 700ms before the bots' turn
+started animating. A seated player's own board is drawn from their authoritative `Frame`, never from
+the queue, so replaying their own move shows them nothing and delays the thing they wanted to watch.
+Own steps are still applied — the delta chain runs through them — but at zero dwell. Automatic phases
+have no player and are unaffected, which is what keeps #18 working in solitaire, where every intent
+is the viewer's own.
+
+### Faster and slower, without a rebuild
+
+`pace` multipliers above 1 are supported and expected — Jesse asked for 2 and 3 — bounded by a new
+`MAX_PACE` of 10 so that `?pace=300` from somebody meaning 3.00 cannot look like a frozen board.
+Every tier scales by the same factor, so **a switching move outlasts an ordinary action at 0.5× and
+at 3× alike**: the relative weighting is the design, and the multiplier is only how fast it runs.
+
+Whole-game animation is now ~5.7 minutes across a 6-day game.
+
+---
+
 ## 0.8.0 — 2026-09-09
 
 **Watching the table.** TODO #13, #15 and #18, which is Gitea#20 steps 2-4 pointed at a seated

docs/plans/jitsi-common-board.md

@@ -361,14 +361,15 @@ Introduce dedicated allow-listed types. Do not derive them with `Omit<Frame, ...
 > **Read those two, not this**, when building steps 2-7. The differences that matter:
 >
 > - **The shape is FLAT, not grouped.** There is no `clock`, `config`, `scoring` or `deckCounts`
->   object. Their contents sit at the top level. **All 37 properties, which is the same list as
+>   object. Their contents sit at the top level. **All 38 properties, which is the same list as
 >   `test/redaction.test.ts`'s allow-list** — `day`, `stage`, `clock` (a time string), `phase`,
 >   `phaseKey`, `actor`, `superintendent`, `deck`, `departments`, `departmentsWhat`,
 >   `departmentDepth`, `salvage`, `yards`, `timetable`, `timetableWhat`, `houseRules`, `mode`,
 >   `optionalRules`, `days`, `minCombinedRevenue`, `maxCollisionsPerDay`, `maxCollisionsTotal`,
->   `collisionsToday`, `collisionsTotal`, `status`, `outcome`, `extraDays`, `extensionVotes`,
+>   `collisionsToday`, `collisionsPrevDay`, `collisionsTotal`, `status`, `outcome`, `extraDays`,
+>   `extensionVotes`,
 >   `official`, `tally`, `players`, `openingRolls`, `trains`, `crewTrays`, `queued`, `division`,
->   `districts`. The first 35 come from `projectSharedTable`; `division` and `districts` are added
+>   `districts`. The first 36 come from `projectSharedTable`; `division` and `districts` are added
 >   by `PublicFrame` itself.
 > - **`protocolVersion` was NOT built** and exists nowhere in the repo. **Decided 2026-09-09: add it
 >   in step 2.** `display.json` carries its own `schemaVersion`, and the SSE wire format is a second,

package.json

@@ -1,6 +1,6 @@
 {
   "name": "station-master",
-  "version": "0.8.0",
+  "version": "0.8.0.3",
   "private": true,
   "type": "module",
   "description": "Station Master — a railroad operations game",

src/engine/advance.ts

@@ -1638,6 +1638,9 @@ function shiftChange(s: GameState, events: GameEvent[]): AdvanceResult {
 
     s.clock.day += 1;
     s.clock.stage = 1;
+    // Captured BEFORE the reset: the Day-end dialog reports the Day that just finished, and it is
+    // drawn from the frame this rollover produces. See `collisionsPrevDay` in `state.ts`.
+    s.collisionsPrevDay = s.collisionsToday;
     s.collisionsToday = 0;
     events.push({ type: 'stageBegan', day: s.clock.day, stage: 1 });
     rotateSeats(s, events);

src/engine/setup.ts

@@ -431,6 +431,7 @@ export function createGame(opts: SetupOptions): GameState {
     turns: freshTurns(playerCount, MOVES_PER_LOCAL_OPS),
     movedThisPhase: new Set(),
     collisionsToday: 0,
+    collisionsPrevDay: 0,
     collisionsTotal: 0,
     status: 'active',
     outcome: null,

src/engine/state.ts

@@ -1119,6 +1119,20 @@ export type GameState = {
   movedThisPhase: Set<TrayId>;
   /** §3.4 — resets at the start of each Day; checked against `config.maxCollisionsPerDay`. */
   collisionsToday: number;
+  /**
+   * What `collisionsToday` held for the Day that just ENDED — captured at the rollover, immediately
+   * before the reset.
+   *
+   * The Day-end dialog exists to report the Day that finished, and it is drawn from the frame AFTER
+   * the rollover, because that is the frame whose `day` went up. So it read `collisionsToday` as 0 no
+   * matter what had happened: Jesse, 2026-09-09, at the end of a Day 1 with two collisions in it —
+   * "it shows a total of two collisions, but zero today ... that does seem to be a contradiction".
+   *
+   * NOT DERIVABLE ON THE CLIENT. A Day turns over inside the phases that run themselves, so in
+   * multiplayer the push that reports the new Day is the same push that reports the reset — a client
+   * may never see the ended Day's final count to remember it.
+   */
+  collisionsPrevDay: number;
   /** §3.4 — never reset; checked against `config.maxCollisionsTotal`. */
   collisionsTotal: number;
   /**

src/sim/pacing.ts

@@ -37,7 +37,9 @@ export type StepKind = 'switching' | 'action' | 'phase' | 'bookkeeping';
  * it is deliberately not the only way to change the pacing. A viewer's own `pace` multiplier
  * (`Settings`, `localStorage`) and a `?pace=` URL parameter both scale these without one, and
  * `pace = 0` turns the animation off entirely, which is also TODO #18's "a player who has seen it a
- * hundred times will want it off".
+ * hundred times will want it off". **Multipliers above 1 are supported and expected** — Jesse asked
+ * for 2 and 3 explicitly after the first play — up to `MAX_PACE`, and every tier scales together so
+ * their relative weighting survives.
  *
  * NOT IN GAME-CREATION SETTINGS, on Jesse's call 2026-09-09: dwell is presentation, not a rule, and
  * `config` rides along in saves and replays. If it ever moves there, the config field supplies this
@@ -46,8 +48,17 @@ export type StepKind = 'switching' | 'action' | 'phase' | 'bookkeeping';
 export const DWELL: Record<StepKind, number> = {
   /** A train physically moving on the board. The thing worth watching, and protected accordingly. */
   switching: 1000,
-  /** A card, a car or a load changing hands somewhere visible. */
-  action: 250,
+  /**
+   * A card, a car or a load changing hands somewhere visible — and the announcement of what a
+   * player is about to do.
+   *
+   * WAS 250ms, WHICH WAS WRONG, and wrong in the way that mattered most: an early-game bot turn has
+   * no switching in it at all, so it was six steps of 250ms and 0ms — **750ms for a whole turn**.
+   * Jesse, from the first real play on `phoenix.local`: *"bot play was way too fast. I briefly saw
+   * that it was the bot's office area then their turn was done."* His instruction had been "start at
+   * 1s and tune down", and that was applied only to switching while this number was invented.
+   */
+  action: 700,
   /**
    * An automatic phase that DID something — TODO #18.
    *
@@ -107,9 +118,17 @@ export function kindOf(cause: StepCause): StepKind {
     case 'redFlag.play':
       return 'action';
 
-    // Ending a phase or a turn, choosing what to do, voting. The consequences are worth watching;
-    // the declaration itself is not, and there are more of these than of anything else.
+    /**
+     * `localOps.choose` IS AN ANNOUNCEMENT, NOT BOOKKEEPING — moved out 2026-09-09 after the first
+     * real play. It is the line that reads "Player Bot 1 chose to SWITCH — six Moves to shunt cars
+     * around the yard": the heading for everything that follows, and at zero dwell nobody ever saw
+     * it, so a bot's turn began with no indication of what it was about to do.
+     */
     case 'localOps.choose':
+      return 'action';
+
+    // Ending a phase or a turn, and voting. Nothing to see: the consequences were the thing, and
+    // there are more of these than of anything else.
     case 'loadUnload.end':
     case 'draw.end':
     case 'switch.end':
@@ -119,9 +138,38 @@ export function kindOf(cause: StepCause): StepKind {
   }
 }
 
-/** How long to show one step, in ms, at a given speed. `pace` of 0 means "do not animate at all". */
+/**
+ * The widest multiplier that is a speed rather than a mistake.
+ *
+ * `pace` has no lower surprise — 0 means off — but an unbounded upper one does: `?pace=300` from
+ * somebody typing 3.00, or a corrupt `localStorage` value, would give a switching move a five-minute
+ * dwell and look exactly like a frozen board. Ten is far beyond any speed anyone would choose (2 and
+ * 3 are the ones actually asked for) and well short of unusable.
+ */
+export const MAX_PACE = 10;
+
+/**
+ * The speeds the on-screen control offers, slowest last.
+ *
+ * `0` is off: every move is drawn at once, as it was before v0.8.0 — TODO #18's "a player who has
+ * seen it a hundred times will want it off". The ladder runs well past 1 because that is what the
+ * first real play asked for: Jesse reached for 7×, and although the `?pace=` he used never took
+ * effect (the splash replaces the query string, so the play page only ever saw `?lobby`), the wish
+ * was real. Watching a bot shunt cars is the point of this feature, and it is worth as long as it
+ * takes.
+ */
+export const PACE_LEVELS = [0, 0.5, 1, 2, 3, 5, 7, 10] as const;
+
+/**
+ * How long to show one step, in ms, at a given speed.
+ *
+ * `pace` scales every tier by the same factor, so **the tiers stay in proportion at any speed** — a
+ * switching move outlasts an ordinary action at 0.5× and at 3× alike. That is deliberate: the
+ * relative weighting is the design (a train moving is worth more attention than a card changing
+ * hands), and the multiplier is only how fast the whole thing runs. `0` means do not animate at all.
+ */
 export function dwellFor(cause: StepCause, pace = 1): number {
-  return Math.round(DWELL[kindOf(cause)] * Math.max(0, pace));
+  return Math.round(DWELL[kindOf(cause)] * Math.min(MAX_PACE, Math.max(0, pace)));
 }
 
 /**
@@ -134,10 +182,25 @@ export function dwellFor(cause: StepCause, pace = 1): number {
  * have to import `DisplayStep` back from the module that imports `StepCause` from it.
  */
 export function dwellForStep(
-  step: { cause: StepCause; lines: readonly unknown[]; frame: { table: object } },
+  step: { cause: StepCause; player: number | null; lines: readonly unknown[]; frame: { table: object } },
   pace = 1,
 ): number {
-  if (step.lines.length > 0) return dwellFor(step.cause, pace);
+  // Off means off, for the clock as much as for anybody's move.
+  if (pace <= 0) return 0;
+  /**
+   * THE SPEED CONTROL IS ABOUT OTHER PEOPLE, NOT ABOUT THE CLOCK.
+   *
+   * A phase keeps its tabled beat at every speed. Measured over 40 turns of a real 3-seat game, the
+   * waiting split almost evenly — 21.0s of other players against 21.0s of phases turning over — so
+   * scaling both put 105 seconds of clock-ticking into a 5× game, all of it after the player's own
+   * move and none of it anything to watch. Jesse, from that game: *"after my turn, when I actually
+   * execute my turn, I'm still subject to that same delay before it moves on. That makes no sense."*
+   *
+   * The phase still gets its beat (TODO #18) — it just does not get longer because somebody wanted
+   * to watch a bot shunt cars.
+   */
+  const speed = step.player === null ? 1 : pace;
+  if (step.lines.length > 0) return dwellFor(step.cause, speed);
   /**
    * A SILENT STEP EARNS A BEAT ONLY WHEN THE CLOCK TURNED OVER — which is TODO #18 exactly: "give
    * every phase a visible beat", for New Train, the Mainline and the shift change.
@@ -146,12 +209,11 @@ export function dwellForStep(
    * silently killed #18: a phase can move trains without saying anything, and those steps were being
    * flashed past. "Anything that changed the board" is wrong the other way: `submit()` steps
    * `advance()` about 4.6 times per intent and most of those merely hand the turn on, so beating on
-   * all of them would cost a quarter of an hour a game. The phase turning over is the thing a player
-   * is being shown, and there are about 180 of those in a full game.
+   * all of them would cost a quarter of an hour a game.
    */
   const table = step.frame.table as Record<string, unknown>;
   const turned = 'phase' in table || 'phaseKey' in table || 'stage' in table || 'day' in table;
-  return turned ? dwellFor(step.cause, pace) : 0;
+  return turned ? dwellFor(step.cause, speed) : 0;
 }
 
 /**

src/sim/view.ts

@@ -437,6 +437,8 @@ export type Frame = {
   maxCollisionsPerDay: number;
   maxCollisionsTotal: number;
   collisionsToday: number;
+  /** What the Day that just ended finished on — see `collisionsPrevDay` in `engine/state.ts`. */
+  collisionsPrevDay: number;
   collisionsTotal: number;
   status: GameState['status'];
   outcome: GameState['outcome'];
@@ -1594,6 +1596,7 @@ export function projectSharedTable(s: GameState) {
     maxCollisionsPerDay: s.config.maxCollisionsPerDay,
     maxCollisionsTotal: s.config.maxCollisionsTotal,
     collisionsToday: s.collisionsToday,
+    collisionsPrevDay: s.collisionsPrevDay,
     collisionsTotal: s.collisionsTotal,
     status: s.status,
     outcome: s.outcome,

src/web/main.ts

@@ -26,6 +26,7 @@ import { createLocalSession, createRemoteSession } from './session.ts';
 import type { PlayerIndex } from '../engine/state.ts';
 import type { PublicDistrict } from '../sim/view.ts';
 import { createStepQueue } from './step-queue.ts';
+import { PACE_LEVELS } from '../sim/pacing.ts';
 import { notice, prefillCode, runLobby } from './lobby.ts';
 import type { LobbyReady } from './lobby.ts';
 import {
@@ -53,6 +54,7 @@ const REMOTE_KEY = 'station-master.remote.v1';
 /** Preset board zoom levels — a fraction applied to the rendered SVG's own pixel dimensions. */
 const ZOOM_LEVELS = [0.75, 1, 1.25, 1.5] as const;
 
+
 /**
  * Small persisted preferences, kept in a `localStorage` key of their own — separate from
  * `SAVE_KEY`. A save is the seed plus the intents and has to stay portable; none of this belongs
@@ -165,7 +167,12 @@ let session: Session;
  * on the next step instead of the next game. `?pace=` wins over the saved setting for this session
  * only.
  */
-const stepQueue = createStepQueue(() => PACE_OVERRIDE ?? settings.pace);
+const stepQueue = createStepQueue(
+  () => PACE_OVERRIDE ?? settings.pace,
+  // Whose moves not to bother replaying — this client's own. Read lazily: `session` is assigned when
+  // a game starts, long after this queue is built.
+  () => (session ? session.seat() : null),
+);
 
 /**
  * Pulls whatever the session has for us into the queue. Called on every push, before rendering.
@@ -194,17 +201,23 @@ function drainIntoQueue(): void {
  * drawn — from `PublicDistrict`, which `officeSvg` can render as-is because it takes board data and
  * has never needed a private viewer.
  */
-function renderWatching(): void {
+function renderWatching(f?: Frame): void {
   const behind = stepQueue.behind();
   const row = $('watching');
-  // Collapsed whenever the board is level with the game — which in solitaire is nearly always, and
-  // between turns in multiplayer too. A row that is always there would be a row nobody reads.
-  if (behind === 0) {
+  /**
+   * VISIBLE WHILE THE BOARD IS BEHIND **OR** STILL SHOWING SOMETHING.
+   *
+   * It used to hide the moment `behind` hit zero — which is the moment the LAST step of a burst goes
+   * up, so the one step a player was most likely to be reading about lost its caption. Collapsed
+   * otherwise: in solitaire that is nearly always, and between turns in multiplayer too, and a row
+   * that is always there is a row nobody reads.
+   */
+  if (behind === 0 && !stepQueue.busy()) {
     row.hidden = true;
     return;
   }
   row.hidden = false;
-  $('watching-behind').textContent = `${behind} behind`;
+  $('watching-behind').textContent = behind === 0 ? 'catching up' : `${behind} behind`;
   /**
    * THE CAPTION IS #15, and this is where that item lands rather than as a line of its own.
    *
@@ -213,8 +226,29 @@ function renderWatching(): void {
    * The queue IS that, so the caption simply names the step being shown, and the counter beside it
    * says how much of the wait is left.
    */
+  /**
+   * WHO, THEN WHAT — Jesse, 2026-09-09: *"it didn't tell me what the actual action was, like who I
+   * was waiting on or what they were doing. I knew I was behind, but I wasn't sure what I was
+   * supposed to be looking for."*
+   *
+   * The caption was there; it was the wrong half of the sentence. Half the waiting is automatic
+   * phases, whose narration reads "Mainline" — accurate, and no answer at all to "who am I waiting
+   * on". So the name goes first, and a phase says so in as many words rather than leaving the reader
+   * to infer that nobody is acting.
+   *
+   * The narrated line is used as it stands otherwise, because `record()` already prefixes it with the
+   * player — "Player Bot 1 moved Train 3 (−1,−2) → (−1,1)" — so a second name would stutter.
+   */
   const showing = stepQueue.showing();
-  $('watching-what').textContent = showing?.lines[0]?.text ?? '';
+  const said = showing?.lines[0]?.text ?? '';
+  const who =
+    showing === null || showing === undefined
+      ? ''
+      : showing.player === null
+        ? 'The Division'
+        : (f?.players[showing.player]?.name ?? `Seat ${seatLabel(showing.player)}`);
+  // A player action already names its actor; a phase does not, so it is introduced.
+  $('watching-what').textContent = showing?.player === null && said !== '' ? `${who}: ${said}` : said;
   /**
    * SKIP COSTS THE ANIMATION AND NEVER THE INFORMATION.
    *
@@ -1257,7 +1291,7 @@ function render(): void {
 
   renderTurnChart(f);
   renderPresence(f);
-  renderWatching();
+  renderWatching(f);
   $('revenue').textContent = String(f.revenue);
   /**
    * THE OBJECTIVE, WITHOUT THE COMMENTARY.
@@ -2598,6 +2632,52 @@ function runSolitaireSetup(params: URLSearchParams, hasSave = false, live: Frame
   dealBtn.onclick = () => commitNewGame(ss, seedField?.value ?? '');
 }
 
+/**
+ * PLAYBACK SPEED — v0.8.0.3, TODO #13.
+ *
+ * Persisted per viewer in `Settings`, so it survives the navigation that was eating `?pace=`. The
+ * queue reads `settings.pace` through a closure on every step, so a change here takes effect on the
+ * very next move rather than the next game.
+ */
+const paceSlowerBtn = document.getElementById('paceslower') as HTMLButtonElement | null;
+const paceFasterBtn = document.getElementById('pacefaster') as HTMLButtonElement | null;
+const paceLabel = document.getElementById('pacelabel');
+if (paceSlowerBtn && paceFasterBtn && paceLabel) {
+  const nearestPace = (): number => {
+    // A saved or URL value need not be on the ladder — `?pace=7` and a hand-edited setting are both
+    // legitimate — so the buttons step from whichever preset is closest rather than refusing to move.
+    const want = PACE_OVERRIDE ?? settings.pace;
+    return PACE_LEVELS.reduce((best, p) => (Math.abs(p - want) < Math.abs(best - want) ? p : best), PACE_LEVELS[0]);
+  };
+  const paintPace = (): void => {
+    const p = PACE_OVERRIDE ?? settings.pace;
+    paceLabel.textContent = p === 0 ? 'off' : `${p}×`;
+    const i = PACE_LEVELS.indexOf(nearestPace() as (typeof PACE_LEVELS)[number]);
+    paceSlowerBtn.disabled = i >= PACE_LEVELS.length - 1;
+    paceFasterBtn.disabled = i <= 0;
+    // A `?pace=` in the URL wins over the setting, so say so rather than showing dead buttons.
+    if (PACE_OVERRIDE !== null) {
+      paceSlowerBtn.disabled = true;
+      paceFasterBtn.disabled = true;
+      paceLabel.textContent = `${PACE_OVERRIDE}× (URL)`;
+    }
+  };
+  const stepPace = (by: number): void => {
+    const i = PACE_LEVELS.indexOf(nearestPace() as (typeof PACE_LEVELS)[number]);
+    const next = PACE_LEVELS[Math.min(PACE_LEVELS.length - 1, Math.max(0, i + by))];
+    if (next === undefined) return;
+    saveSettings({ pace: next });
+    paintPace();
+    // The row's countdown is measured in steps that will dwell, so a change to 0 empties it at once.
+    renderWatching();
+  };
+  // Slower is a BIGGER multiplier, so "−" walks up the ladder. Labelled by what it does to the game,
+  // not to the number: a player pressing "slower" wants to watch for longer.
+  paceSlowerBtn.onclick = () => stepPace(1);
+  paceFasterBtn.onclick = () => stepPace(-1);
+  paintPace();
+}
+
 const zoomOutBtn = document.getElementById('zoomout') as HTMLButtonElement | null;
 const zoomInBtn = document.getElementById('zoomin') as HTMLButtonElement | null;
 const zoomLabel = document.getElementById('zoomlabel');

src/web/panels.ts

@@ -180,7 +180,7 @@ export function dayEndHtml(f: Frame): string {
     ahead +
     standingsHtml(f) +
     targetHtml(f) +
-    collisionsHtml(f)
+    collisionsHtml(f, ended)
   );
 }
 
@@ -238,13 +238,28 @@ function targetHtml(f: Frame): string {
  * its config and enforces neither, so reporting a collision budget there would put a rule on
  * screen that this game does not have.
  */
-function collisionsHtml(f: Frame): string {
+function collisionsHtml(f: Frame, endedDay?: number): string {
   const scoredOnCollisions =
     (f.mode === 'competitive' || f.mode === 'coop') &&
     (f.maxCollisionsTotal > 0 || f.maxCollisionsPerDay > 0);
-  return scoredOnCollisions
-    ? `<p>Collisions: <b>${f.collisionsToday}</b> today, <b>${f.collisionsTotal}</b> in all.</p>`
-    : '';
+  if (!scoredOnCollisions) return '';
+  /**
+   * "TODAY" IS THE WRONG WORD IN A DAY-END DIALOG, and it read as a contradiction.
+   *
+   * That dialog is drawn from the frame whose `day` went UP — which is the same frame in which
+   * `collisionsToday` was reset — so it reported 0 however many there had been. Jesse, 2026-09-09,
+   * at the end of a Day 1 with two collisions in it: "it shows a total of two collisions, but zero
+   * today ... that does seem to be a contradiction."
+   *
+   * So when the caller knows which Day just ended it says so by name, and reads the count captured at
+   * the rollover. The end-of-game results screen passes nothing and keeps "today", where the Day has
+   * not turned over and the word is accurate.
+   */
+  const [count, when] =
+    endedDay === undefined
+      ? [f.collisionsToday, 'today']
+      : [f.collisionsPrevDay, `on Day ${endedDay}`];
+  return `<p>Collisions: <b>${count}</b> ${when}, <b>${f.collisionsTotal}</b> in all.</p>`;
 }
 
 /**

src/web/play.html

@@ -125,6 +125,7 @@ section{background:var(--panel);border:1px solid var(--line);border-radius:7px;
 .wbehind{font-variant-numeric:tabular-nums;font-weight:700;color:#c9cee0;
   background:#22263a;border:1px solid #343a52;border-radius:10px;padding:1px 8px;white-space:nowrap}
 #watching-skip{font-size:11px;padding:2px 10px;border-color:#e0b060;color:#e0b060;flex:none}
+#watching-who{color:#c9cee0;font-weight:700}
 #presence:empty{display:none}
 /* division strip */
 #division{display:flex;gap:7px;overflow-x:auto;padding-bottom:4px}
@@ -872,6 +873,14 @@ ul.blocked li{padding:2px 0}
   <span class="zoom" title="Zoom the Division map and your Office Area. Both already scroll — this only changes their size.">
     <button id="zoomout" aria-label="Zoom out">−</button><span id="zoomlabel">100%</span><button id="zoomin" aria-label="Zoom in">+</button>
   </span>
+  <!-- HOW FAST OTHER PLAYERS' TURNS PLAY BACK — v0.8.0.3, TODO #13.
+       A CONTROL RATHER THAN ONLY A URL PARAMETER. `?pace=` shipped first and is unreachable through
+       the front door: `index.html`'s two doors are `play.html?lobby` and `play.html?solitaire`, so
+       arriving from the splash REPLACES the query string and any pace with it. Jesse played a whole
+       game believing he was at 7x when he was at 1x. -->
+  <span class="zoom" title="How long another player's or a bot's move is held on screen before the next one. Yours are never delayed. Off draws every move at once, as it did before v0.8.0.">
+    <button id="paceslower" aria-label="Slower playback">−</button><span id="pacelabel">1×</span><button id="pacefaster" aria-label="Faster playback">+</button>
+  </span>
   <button id="undo" title="Take the last action back. The save is the seed plus the moves made, so this replays the game without the last one — as far back as you like.">Undo</button>
   <button id="savefile" title="Download this game as a save file you can replay or share">Save replay</button>
   <button id="newgame" title="Set up a fresh game — the seed, the table, the opening hand and what the three economies pay. Opens the same screen a new solitaire game starts from, with your current rules filled in; your game in progress is kept until you press Deal, and Continue puts it straight back.">New game</button>
@@ -906,9 +915,12 @@ ul.blocked li{padding:2px 0}
      and Skip. Empty and collapsed whenever the board is level with the game, which in solitaire is
      almost always. -->
 <div id="watching" hidden>
+  <!-- SKIP FIRST, on the left. It sat on the far right and a player's eye is on the countdown, not at
+       the other end of the row — Jesse, 2026-09-09: "the skip button should be on the far left, in
+       front of where it says [the count], so it's always close to where people are looking." -->
+  <button id="watching-skip" class="ghost" type="button" title="Stop animating and jump the board to where the game actually is. Nothing is lost — every line is already in the History panel.">Skip</button>
   <span id="watching-behind" class="wbehind"></span>
   <span id="watching-what"></span>
-  <button id="watching-skip" class="ghost" type="button" title="Stop animating and jump the board to where the game actually is. Nothing is lost — every line is already in the History panel.">Skip</button>
 </div>
 
 <main>

src/web/splash.ts

@@ -40,6 +40,29 @@ if (heroImage && lightbox) {
  * other way costs a click and a lobby that says it cannot reach a server — which is legible, and
  * recoverable. So a slow or flaky probe leaves the door alone; only a definite answer closes it.
  */
+/**
+ * CARRY `?pace=` THROUGH THE DOORS — v0.8.0.3.
+ *
+ * Both doors are static hrefs that REPLACE the query string (`play.html?lobby`,
+ * `play.html?solitaire`), so a `pace` typed on this page was silently dropped on the way in: Jesse
+ * played a whole game believing he was at 7× when the play page had only ever seen `?lobby`. The
+ * durable answer is the speed control on the play screen, which persists per viewer — this keeps the
+ * URL lever honest for handing two playtesters different speeds, which is the only thing it was ever
+ * for.
+ */
+try {
+  const pace = new URLSearchParams(location.search).get('pace');
+  if (pace !== null) {
+    for (const door of Array.from(document.querySelectorAll('a.door'))) {
+      const href = door.getAttribute('href');
+      // Only the doors into the game, and only ones that have not been disabled above.
+      if (href?.startsWith('./play.html?')) door.setAttribute('href', `${href}&pace=${encodeURIComponent(pace)}`);
+    }
+  }
+} catch {
+  // A door that keeps its own href is the status quo, not a broken page.
+}
+
 const mpDoor = document.getElementById('door-multiplayer');
 if (mpDoor) {
   const close = (): void => {

src/web/step-queue.ts

@@ -47,14 +47,32 @@ export type StepQueue = {
   busy(): boolean;
 };
 
-/** `pace` is read on every step rather than captured, so changing the setting takes effect at once. */
-export function createStepQueue(pace: () => number = () => 1): StepQueue {
+/**
+ * `pace` is read on every step rather than captured, so changing the setting takes effect at once.
+ *
+ * `viewer` says which seat is watching, so THIS PLAYER'S OWN MOVES COST NO TIME. They are already on
+ * screen: a seated player's own board is drawn from their authoritative `Frame`, not from the queue,
+ * so holding their click for a dwell shows them nothing and delays the thing they actually want to
+ * watch — the 700ms before a bot's turn starts animating is 700ms of their own move being replayed
+ * at them. The step is still APPLIED, because the delta chain runs through it.
+ *
+ * Automatic phases have no player and are unaffected, which is what keeps TODO #18 working in
+ * solitaire where every intent is the viewer's own.
+ */
+export function createStepQueue(
+  pace: () => number = () => 1,
+  viewer: () => number | null = () => null,
+): StepQueue {
   let shown: PublicFrame | null = null;
   let last: DisplayStep | null = null;
   let pending: DisplayStep[] = [];
   /** When the step now on screen is due to give way. Null when nothing is waiting. */
   let dueAt: number | null = null;
 
+  /** How long this step holds the screen — zero for the viewer's own moves; see above. */
+  const dwell = (step: DisplayStep): number =>
+    step.player !== null && step.player === viewer() ? 0 : dwellForStep(step, pace());
+
   /** Applies one step to the displayed board. A step's delta chains off the previous step's frame. */
   const show = (step: DisplayStep): void => {
     shown = applyPublicDelta(shown, step.frame);
@@ -76,7 +94,10 @@ export function createStepQueue(pace: () => number = () => 1): StepQueue {
 
     advance(now) {
       if (pending.length === 0) {
-        dueAt = null;
+        // The LAST step of a burst still owes its dwell. Clearing `dueAt` here reported the queue
+        // idle the instant that step was shown, which snapped the district panel home before anyone
+        // could look at it — see `busy()`.
+        if (dueAt !== null && now >= dueAt) dueAt = null;
         return false;
       }
       // First step of a burst: show it immediately rather than waiting out a dwell for a board the
@@ -84,7 +105,7 @@ export function createStepQueue(pace: () => number = () => 1): StepQueue {
       if (dueAt === null) {
         const first = pending.shift()!;
         show(first);
-        dueAt = now + dwellForStep(first, pace());
+        dueAt = now + dwell(first);
         return true;
       }
       let drew = false;
@@ -97,7 +118,7 @@ export function createStepQueue(pace: () => number = () => 1): StepQueue {
       while (pending.length > 0 && now >= dueAt) {
         const next = pending.shift()!;
         show(next);
-        dueAt = dueAt + dwellForStep(next, pace());
+        dueAt = dueAt + dwell(next);
         drew = true;
       }
       if (pending.length === 0 && now >= dueAt) dueAt = null;
@@ -113,8 +134,19 @@ export function createStepQueue(pace: () => number = () => 1): StepQueue {
     },
 
     current: () => shown,
-    behind: () => pending.filter((s) => dwellForStep(s, pace()) > 0).length,
+    behind: () => pending.filter((s) => dwell(s) > 0).length,
     showing: () => last,
-    busy: () => pending.length > 0,
+    /**
+     * STILL SHOWING SOMETHING, not just still holding something back.
+     *
+     * This was `pending.length > 0`, which went false the moment the last step of a burst was
+     * shown — so the animation loop stopped and the district panel snapped back to the viewer's own
+     * board without that step ever being visible. Reported from real play: "I briefly saw that it was
+     * the bot's office area, then their turn was done and it pointed back to my office area."
+     *
+     * `dueAt` is non-null exactly while the step on screen has time left, so the two together mean
+     * "there is more to come, or what is up has not had its moment yet".
+     */
+    busy: () => pending.length > 0 || dueAt !== null,
   };
 }

test/pacing.test.ts

@@ -14,7 +14,7 @@ import { readFileSync } from 'node:fs';
 import { dirname, join } from 'node:path';
 import { fileURLToPath } from 'node:url';
 
-import { DWELL, dwellFor, dwellForStep, kindOf, watchableCount } from '../src/sim/pacing.ts';
+import { DWELL, MAX_PACE, PACE_LEVELS, dwellFor, dwellForStep, kindOf, watchableCount } from '../src/sim/pacing.ts';
 import type { StepKind } from '../src/sim/pacing.ts';
 import type { Intent } from '../src/engine/intents.ts';
 
@@ -50,7 +50,13 @@ describe('pacing — dwell by kind', () => {
     assert.equal(kindOf('draw.end'), 'bookkeeping');
     assert.equal(kindOf('loadUnload.end'), 'bookkeeping');
     assert.equal(kindOf('switch.end'), 'bookkeeping');
-    assert.equal(kindOf('localOps.choose'), 'bookkeeping');
+    /**
+     * `localOps.choose` IS AN ANNOUNCEMENT, not bookkeeping — moved 2026-09-09 after the first real
+     * play on `phoenix.local`. It is the line reading "Player Bot 1 chose to SWITCH", the heading for
+     * everything that follows, and at zero dwell a bot's turn began with no sign of what it was about
+     * to do.
+     */
+    assert.equal(kindOf('localOps.choose'), 'action');
 
     assert.ok(DWELL.switching > DWELL.action, 'switching must outrank an ordinary action');
     assert.equal(DWELL.bookkeeping, 0, 'bookkeeping must cost the player no time at all');
@@ -63,6 +69,30 @@ describe('pacing — dwell by kind', () => {
     assert.equal(dwellFor('switch.move'), 1000);
   });
 
+  it('supports multipliers above 1, and keeps the tiers in proportion at every speed', () => {
+    /**
+     * Jesse, 2026-09-09, after the first play: keep switching and ordinary actions at DIFFERENT
+     * delays, and support 2.0 and 3.0 as well as 1.5. So this pins both halves — that the larger
+     * multipliers work at all, and that scaling never flattens the tiers into each other, since the
+     * relative weighting is the design and the multiplier is only how fast it runs.
+     */
+    for (const pace of [0.5, 1, 1.5, 2, 3]) {
+      assert.equal(dwellFor('switch.move', pace), Math.round(DWELL.switching * pace));
+      assert.equal(dwellFor('card.play', pace), Math.round(DWELL.action * pace));
+      assert.ok(
+        dwellFor('switch.move', pace) > dwellFor('card.play', pace),
+        `at ${pace}x a switching move no longer outlasts an ordinary action`,
+      );
+      assert.equal(dwellFor('draw.end', pace), 0, 'bookkeeping stays free at every speed');
+    }
+    // A whole switching exercise at 3x is slow on purpose, and still not absurd.
+    assert.equal(dwellFor('switch.move', 3) * 6, 18_000);
+
+    // And a typo cannot freeze the board: ?pace=300 from somebody meaning 3.00.
+    assert.equal(dwellFor('switch.move', 300), DWELL.switching * MAX_PACE);
+    assert.equal(dwellFor('switch.move', MAX_PACE + 5), dwellFor('switch.move', MAX_PACE));
+  });
+
   it('scales with the viewer\'s pace, and 0 turns it off', () => {
     assert.equal(dwellFor('switch.move', 1), 1000);
     assert.equal(dwellFor('switch.move', 0.5), 500);
@@ -90,6 +120,31 @@ describe('pacing — dwell by kind', () => {
     assert.equal(watchableCount(queue, 0), 0, 'with animation off, nothing is behind');
   });
 
+  it('offers speeds a player actually reached for, and none the code would clamp', () => {
+    /**
+     * Jesse played a whole game believing he was at 7× and was in fact at 1×: `?pace=` shipped as the
+     * only lever, and `index.html`'s doors are `play.html?lobby` / `play.html?solitaire`, so arriving
+     * from the splash REPLACES the query string. Hence a real control on the play screen, and hence
+     * this ladder — which must reach the speeds people ask for and must not offer one that
+     * `dwellFor` would silently clamp.
+     */
+    assert.equal(PACE_LEVELS[0], 0, 'off must be the first rung — #18 wants it turned off');
+    assert.ok(PACE_LEVELS.includes(1), 'the default must be on the ladder');
+    assert.ok(PACE_LEVELS.includes(7), '7x was asked for by name');
+    for (const p of PACE_LEVELS) {
+      assert.ok(p <= MAX_PACE, `${p}x is past MAX_PACE, so the control would lie about it`);
+      assert.equal(dwellFor('switch.move', p), Math.round(DWELL.switching * p));
+    }
+    // Strictly increasing, so stepping the control always changes the speed.
+    for (let i = 1; i < PACE_LEVELS.length; i++) {
+      assert.ok(PACE_LEVELS[i]! > PACE_LEVELS[i - 1]!, 'the ladder must be strictly increasing');
+    }
+    // The slowest rung has to be slow enough to be worth having: six switching moves at the top of
+    // the ladder is a full minute, which is the "watch them struggle" case.
+    const slowest = dwellFor('switch.move', PACE_LEVELS[PACE_LEVELS.length - 1]!) * 6;
+    assert.ok(slowest >= 60_000, `the slowest a switching turn can be watched is ${slowest}ms`);
+  });
+
   it('a silent step beats only when the clock turns over — TODO #18', () => {
     /**
      * Both obvious rules were wrong, so both are pinned. "No narration, no dwell" flashed past
@@ -97,28 +152,69 @@ describe('pacing — dwell by kind', () => {
      * that changed the board" beat on every turn hand-off — `submit()` steps `advance()` about 4.6
      * times per intent — which came to a quarter of an hour a game.
      */
-    const silent = { cause: 'phase' as const, lines: [] as string[] };
+    const silent = { cause: 'phase' as const, player: null, lines: [] as string[] };
     assert.equal(dwellForStep({ ...silent, frame: { table: { actor: 2 } } }), 0, 'a turn hand-off shows nothing');
     assert.equal(dwellForStep({ ...silent, frame: { table: {} } }), 0, 'a step that changed nothing shows nothing');
     assert.equal(dwellForStep({ ...silent, frame: { table: { phase: 'mainline' } } }), DWELL.phase);
     assert.equal(dwellForStep({ ...silent, frame: { table: { stage: 4 } } }), DWELL.phase);
     // Narration always earns the dwell of whatever caused it, clock or no clock.
     assert.equal(
-      dwellForStep({ cause: 'switch.move', lines: ['moved'], frame: { table: {} } }),
+      dwellForStep({ cause: 'switch.move', player: 1, lines: ['moved'], frame: { table: {} } }),
       DWELL.switching,
     );
   });
 
+  it('the speed control stretches other people, not the clock', () => {
+    /**
+     * Jesse, from a real 5× game: *"after my turn, when I actually execute my turn, I'm still subject
+     * to that same delay before it moves on. That makes no sense."* It was not his move being
+     * replayed — it was the automatic phases behind it, which were scaling with `pace` along with
+     * everything else. Measured over 40 turns, the waiting split almost evenly between other players
+     * and phases turning over, so a 5× game spent 105 seconds on the clock alone.
+     */
+    const phase = { cause: 'phase' as const, player: null, lines: ['New Train'], frame: { table: { phase: 'newTrain' } } };
+    const theirs = { cause: 'switch.move' as const, player: 1, lines: ['moved'], frame: { table: {} } };
+    for (const pace of [1, 3, 5, 7]) {
+      assert.equal(dwellForStep(phase, pace), DWELL.phase, `a phase beat grew at ${pace}x`);
+      assert.equal(dwellForStep(theirs, pace), DWELL.switching * pace);
+    }
+    // Off still means off, for the clock as much as for anybody's move.
+    assert.equal(dwellForStep(phase, 0), 0);
+    assert.equal(dwellForStep(theirs, 0), 0);
+  });
+
   it('a real switching turn is watchable in a few seconds, not tens of them', () => {
     // Six moves is the engine's cap per crew ("N of 6 Moves left"), so this is the worst ordinary
-    // case for one crew and the arithmetic the design promised: ~6s to watch a whole exercise.
+    // case for one crew: the announcement, six moves, and an end that shows nothing.
     const turn: Intent['type'][] = [
       'localOps.choose',
       ...Array<Intent['type']>(6).fill('switch.move'),
       'switch.end',
     ];
     const total = turn.reduce((ms, i) => ms + dwellFor(i), 0);
-    assert.equal(total, 6000);
-    assert.equal(watchableCount(turn), 6, 'the choose and the end are not things to watch');
+    assert.equal(total, DWELL.action + 6 * DWELL.switching);
+    assert.ok(total > 5_000 && total < 10_000, `a switching turn takes ${total}ms to watch`);
+    assert.equal(watchableCount(turn), 7, 'the six moves and the announcement; not the end');
+  });
+
+  it("a bot's ordinary turn is followable, which is what the first real play was not", () => {
+    /**
+     * MEASURED FROM A REAL GAME, then pinned. Jesse, after installing v0.8.0 on `phoenix.local`:
+     * *"bot play was way too fast. I briefly saw that it was the bot's office area then their turn
+     * was done."* This is the shape that turn actually had — no switching in it at all, because
+     * switching is not legal until there is track down — and under the original values it came to
+     * 750ms for the whole thing.
+     */
+    const turn: Intent['type'][] = [
+      'localOps.choose',
+      'draw.fromHomeOffice',
+      'card.play',
+      'draw.end',
+      'localOps.choose',
+      'freightAgent.stockOutbound',
+    ];
+    const total = turn.reduce((ms, i) => ms + dwellFor(i), 0);
+    assert.ok(total >= 3_000, `an ordinary bot turn is only ${total}ms — too fast to follow`);
+    assert.equal(watchableCount(turn), 5, 'only the turn-ending bookkeeping is free');
   });
 });

test/redaction.test.ts

@@ -524,6 +524,9 @@ describe('#91 — nothing private survives serialisation, in any state', () => {
       // The rules the game was dealt under, and the score.
       'houseRules', 'mode', 'optionalRules', 'days', 'minCombinedRevenue',
       'maxCollisionsPerDay', 'maxCollisionsTotal', 'collisionsToday', 'collisionsTotal',
+      // What the Day that just ended finished on. Public for the same reason the running counts are:
+      // a collision happens on the Mainline in front of everybody.
+      'collisionsPrevDay',
       'status', 'outcome', 'extraDays', 'extensionVotes', 'official', 'tally',
       // Names, seats, revenue and HAND SIZE — never hand contents.
       'players',

test/step-queue.test.ts

@@ -87,7 +87,8 @@ describe('the step queue', () => {
     q.reset(baseline(1917398));
 
     // Only the bookkeeping: it must all collapse into a single advance.
-    const bookkeeping = steps.filter((s) => s.cause.endsWith('.end') || s.cause === 'localOps.choose');
+    // `.end` only: `localOps.choose` became an announcement worth watching after the first real play.
+    const bookkeeping = steps.filter((s) => s.cause.endsWith('.end'));
     assert.ok(bookkeeping.length > 10, 'not enough bookkeeping steps to prove the collapse');
     q.push(bookkeeping);
     q.advance(0);
@@ -173,6 +174,62 @@ describe('the step queue', () => {
     assert.equal(half.behind(), 1, 'at half pace, half the dwell should have advanced one step');
   });
 
+  it('holds the LAST step of a burst for its dwell — the v0.8.0 snap-back bug', () => {
+    /**
+     * REGRESSION. `busy()` was `pending.length > 0`, so the instant the final step of a burst was
+     * shown the queue reported idle: the animation loop stopped and the district panel snapped back
+     * to the viewer's own board without that step ever being looked at. Jesse, from the first real
+     * play on `phoenix.local`: *"I briefly saw that it was the bot's office area then their turn was
+     * done and it pointed back to my office area"*, and the countdown row appeared "very briefly".
+     *
+     * The panel follows `busy()`, so this is the property that keeps somebody else's board on screen
+     * for as long as their move is being shown.
+     */
+    const { steps } = realSteps(1917398, 400);
+    const one = steps.filter((s) => s.cause === 'switch.move').slice(0, 1);
+    assert.equal(one.length, 1);
+
+    const q = createStepQueue();
+    q.reset(baseline(1917398));
+    q.push(one);
+
+    q.advance(0);
+    assert.equal(q.behind(), 0, 'nothing is queued behind it');
+    assert.equal(q.busy(), true, 'but it is still being shown, so the queue is not idle');
+
+    q.advance(DWELL.switching - 1);
+    assert.equal(q.busy(), true, 'still inside its dwell');
+
+    q.advance(DWELL.switching);
+    assert.equal(q.busy(), false, 'and idle only once its moment has passed');
+  });
+
+  it("does not spend time replaying the viewer's own moves", () => {
+    // A seated player's own board is drawn from their authoritative Frame, so they have already seen
+    // their own click. Holding it delays the thing they wanted to watch — a bot's turn.
+    const { steps } = realSteps(1917398, 400);
+    const mine = steps.filter((s) => s.player === 0 && s.cause === 'switch.move').slice(0, 3);
+    assert.equal(mine.length, 3, 'need three of seat 0\'s own moves');
+
+    const asSeat0 = createStepQueue(() => 1, () => 0);
+    asSeat0.reset(baseline(1917398));
+    asSeat0.push(mine);
+    // Twice at the same instant: the first call shows the head of the burst, the second collapses the
+    // zero-dwell run behind it. In the page that is two animation frames, ~16ms apart.
+    asSeat0.advance(0);
+    asSeat0.advance(0);
+    assert.equal(asSeat0.busy(), false, "the viewer's own moves must cost no time at all");
+    assert.equal(asSeat0.behind(), 0, 'and must never be counted as something to wait for');
+
+    // The same steps seen by somebody else are worth watching.
+    const asSpectator = createStepQueue(() => 1, () => 1);
+    asSpectator.reset(baseline(1917398));
+    asSpectator.push(mine);
+    asSpectator.advance(0);
+    assert.equal(asSpectator.busy(), true, "another seat's moves are worth showing");
+    assert.equal(asSpectator.behind(), 2);
+  });
+
   it('a reset discards the backlog rather than merging it onto a new baseline', () => {
     /**
      * A reconnecting client holds steps whose deltas chain off a baseline the server has moved past.

test/web.test.ts

@@ -3614,6 +3614,46 @@ describe('the Day rolling over says so (Gitea#10)', () => {
     assert.ok(html.includes('3 Days left'), `the Days remaining are wrong:\n${html}`);
   });
 
+  it('reports the ENDED Day\'s collisions, not the fresh Day\'s zero', () => {
+    /**
+     * Jesse, 2026-09-09, at the end of a Day 1 with two collisions in it: *"It shows a total of two
+     * collisions, but zero today. Since we just finished day one, that does seem to be a
+     * contradiction."*
+     *
+     * The cause is a one-line ordering fact: `advance.ts` increments the Day and then zeroes
+     * `collisionsToday`, and this dialog is drawn from the frame whose Day went UP — so it read the
+     * fresh Day's zero and printed it beside a running total that could not agree with it. The count
+     * is captured at the rollover now, and the dialog names the Day rather than saying "today".
+     */
+    const s = createEngineGame({
+      id: 'collide',
+      seed: 5,
+      config: {
+        mode: 'competitive',
+        days: 5,
+        minCombinedRevenue: 60,
+        maxCollisionsPerDay: 3,
+        maxCollisionsTotal: 10,
+        pvpCardsAllowed: false,
+        optionalRules: { reducedVisibility: false, employeeRotation: false, emergencyToolbox: false },
+      },
+      playerNames: ['Joe', 'Bot 1'],
+    });
+    // The state as the rollover out of Day 1 leaves it: two collisions happened, `today` is reset.
+    s.clock.day = 2;
+    s.collisionsPrevDay = 2;
+    s.collisionsToday = 0;
+    s.collisionsTotal = 2;
+
+    const html = dayEndHtml(snapshot(s, [], null));
+    assert.ok(html.includes('Day 1 has ended'), `wrong Day named:\n${html}`);
+    assert.ok(html.includes('<b>2</b> on Day 1'), `the ended Day's collisions are wrong:\n${html}`);
+    assert.ok(html.includes('<b>2</b> in all'), `the running total is wrong:\n${html}`);
+    assert.doesNotMatch(html, /<b>0<\/b> today/, `still reporting the fresh Day's zero:\n${html}`);
+    // The contradiction itself: a Day-end dialog must never claim fewer in all than on that Day.
+    assert.doesNotMatch(html, /<b>0<\/b> on Day 1/, 'reported no collisions on a Day that had two');
+  });
+
   it('counts the last Day as the last Day rather than promising more', () => {
     const html = dayEndHtml(frameAt(6));
     assert.ok(html.includes('Day 5 has ended'), 'the final Day is misnamed');