e2e: per-test timescale speedup — FAST/CHAIN rates for pure arrival waits (v1.8.0)

- actions/hover/two-planes: FAST=10_000 for expiry fast-forwards, landing
  waits and fly-home cleanups; queue: + CHAIN=4_000 for auto-chain
  observation, refuel test at explicit 2_000 with relaxed fuel bound
- map-interactions: FAST=50_000 chain hops (~3 min -> ~15 s per direction);
  crossing leg stays at SCALE and now sets it BEFORE dispatch (a leftover
  fast rate would let the dispatch tick consume most of the leg)
- hover fleet-row test asserts flying from the fly endpoint's own snapshot
  and freezes immediately instead of speeding up then freezing mid-leg
- two-planes parallel-flight: both-grounded poll after cleanup dispatches so
  afterEach heal no longer pays the range-edge return leg at SCALE

CHANGELOG.md

@@ -1,5 +1,12 @@
 # Changelog
 
+## v1.8.0 — 2026-09-27 20:13
+- **E2E speedup (test-only, no game code changed).** Every Playwright spec that fast-forwards legs now uses the fastest rate its assertions allow instead of a blanket `SCALE = 1000`: pure arrival waits (no mid-leg observation) run at `FAST` — a lagged tick can only land a plane sooner, never skip an observed state. Expected CI wall time for the affected tests drops ~5–6× overall (e.g. the antimeridian test's relocation chains go from ~3 min to ~15 s per direction; two-planes' parallel-flight test loses its ~24 s afterEach heal leg).
+- **Per-test rates.** `actions.spec.ts` + `hover.spec.ts` + `two-planes.spec.ts`: new local `FAST = 10_000` for expiry fast-forwards, landing waits and fly-home cleanups. `queue.spec.ts`: same `FAST`, plus `CHAIN = 4_000` for the auto-chain observation (max per-tick advance 2 s × S stays below the LEMD-length leg's ~1e7 sim-ms duration so one lagged tick can't complete it before a poll sees the chained state); its refuel-before-chained-leg test moves to an explicit `2_000` with the fuel lower bound relaxed from `> 1.3×required` to `> required` (at 2_000× the snapshot can land ~0.4×required into leg 2's burn under dind load; a wrong top-up level still lands far below the bound, and the exact level stays pinned by the unchanged cash assertion). `map-interactions.spec.ts`: `FAST = 50_000` for chain hops (≤3400 km each) and arrival waits.
+- **Kept at SCALE = 1000 where mid-leg observation is real.** The antimeridian crossing leg itself (freshness gate ~700 km ≈ 5 s of flight + >3 fresh samples assume ~14 s of real-time flight), the queue spec's final-hop chain window (EHAM→EDDF is only 2_587 sim-s — at higher rates a `/api/state` 500-retry gap could advance past it), and map-interactions' mid-flight click window. The antimeridian crossing dispatch now sets SCALE BEFORE the fly POST: with FAST-rate chains, a leftover fast rate would let the dispatch tick itself consume most of the leg before the setScale lands (flake path).
+- **hover.spec.ts fleet-row test restructured.** It no longer speeds up to 1000× and then freezes mid-leg; it asserts `flying` from the fly endpoint's OWN post-action snapshot (no poll race) and freezes at scale 1 immediately — same pattern as its package-hover sibling, so the ~10 s leg can't land under the assertions.
+- **two-planes parallel-flight test.** Landing waits move to FAST, and a new both-grounded poll after the cleanup dispatches lets the range-edge return legs finish while the clock is still at FAST — previously afterEach's `healPlaneHome` paid that ~24 s leg back at SCALE.
+
 ## v1.7.0 — 2026-09-26 20:24
 - **Multiplayer step 12/12 (ISS-116): multiplayer test coverage.** The final campaign step closes the two remaining verification gaps with tests only — no production game code changed: the #135 freeze rule (`shouldAdvanceTime`) already counts ALL planes regardless of owner, and the mint hook has supported a username override since v1.1.0 (the e2e helper just never forwarded it).
 - **E2E.** New `e2e/multiplayer.spec.ts` (serial) plays two live sessions: player A is the suite's usual local player; player B is a second account minted through the same hook with a username override. Test 1 verifies the N-player clock end-to-end (#135 × ISS-113): while A flies and B stays grounded, the world clock advances on every read (one owner airborne is enough to keep time moving), then after A lands — EVERY player grounded — the clock freezes exactly across ~2.5 s of real idleness. Test 2 verifies ownership isolation: both sessions see the same shared fleet with `ownerId`/`ownerName`, A gets 403 "not your plane" on B's starter (fly AND sell — the 403 must beat B's last-plane 409), B gets 403 on A's, a grant to local moves only A's account row while B's stays untouched, and both sessions see the v1.6.0 `players` summary with per-owner plane counts.