Correct misleading crash-recovery comment (release pair)

[ryanbr]

Comment-only docfix on the release pair. No behavioral change, no version bump (per the no-per-PR-version convention) — bullet added under ## Unreleased for the next roll-up.

What the old comment said (false)

The IVS WASM worker error handler in vaft.user.js + vaft-ublock-origin.js previously claimed:

"…and existing reload cooldown prevents runaway restart loops."

That's verifiably false. The crash handler calls doTwitchPlayerTask(false, true, 'early') directly on the main thread. The reload-cooldown logic (tooSoonSinceLastReload, escalation, 30s/90s) lives in the worker's processM3U8 reload-decision path — which crash-recovery never traverses. doTwitchPlayerTask itself has no throttle. So a hypothetical persistently-crashing WASM would currently produce an unbounded crash → uncooled hard reload → fresh worker → crash loop, not a cooldown-throttled one.

This misleading comment was what originally pulled me into the circuit-breaker thread (closed PR #230). Fixing the comment regardless removes the documented-but-absent safety property trap.

What the new comment says (accurate)

• Lists all three observed WASM-error variants: index out of bounds, indirect call signature mismatch, indirect call to null (the v82 and v83 logs gave us the latter two).
• States the per-worker crashed flag dedupes the multi-event burst from one crash — and is per-closure (does NOT survive the reload).
• States honestly: no cross-worker storm guard is implemented here. Intentional — observed crashes have all been single + recoverable.
• Notes the trip-wire for revisiting: clustered crashes ≥2 within ~60s in one session. If that ever appears in field logs, the circuit-breaker design from closed PR v68.4.0: circuit-breaker for IVS WASM crash recovery #230 is the ready answer.

Scope

• vaft.user.js + vaft-ublock-origin.js: comment block rewritten (6 lines → 10 lines, both files synced).
• CHANGELOG.md: bullet under ## Unreleased → ### Documentation.
• Testing pair NOT touched. The testing variant (from the v651 testing-direct work during v68.4.0: circuit-breaker for IVS WASM crash recovery #230, which was never reverted) already has the full circuit-breaker code + its own accurate comment. Leaving that as intentional release-vs-testing divergence, per CLAUDE.md "Testing files include experimental features."

acorn-clean on both release files. Single squashed commit.

🤖 Generated with Claude Code