[Moltenship]

Outcome

Part of #263

The 1.0.0-rc.2 release gate failed in Firefox (run 38763042004): #261's new drag-copy test timed out there in all four variants. No local run can reproduce it on macOS 27 (#182). The owner wants the release candidate out first, so vp run browsers leaves Firefox out until #263 brings it back, with its cause found.

Contract and scope

• Public API or behavior changed: none. Only the test runs change.
• Packages/apps/registry items affected: tools/a11y (its browsers mode), AGENTS.md, docs/internal/quality-gates.md.
• Explicitly out of scope: the Firefox failure itself (Add the Firefox tests back #263). CI still installs Firefox, and its job keeps the name "Firefox and WebKit", which the workflow tests pin.
• ADR added or updated, if required: none.

Acceptance criteria

• The PR closes one primary implementation task: it is part of Add the Firefox tests back #263, which stays open.
• Native blocker relationships still match the real merge order.
• Public source is imported through package entry points; there are no deep imports.

Evidence

• vp check
• vp run ready:full passes locally. On macOS 27 it ran WebKit alone before this change too.
• CI's on-demand release-gate and browsers jobs on this branch: https://github.com/Moltenship/time/actions/runs/38767364580
• N/A SSR, performance, visual, a11y and registry evidence: no product code changes.

Architecture invariants

• No product code changes.

Risk and rollout

• Main risk: Firefox regressions go unnoticed until Add the Firefox tests back #263.
• Safe rollback/reversal: set firefoxRuns back in tools/a11y/vite.config.ts.
• Follow-up work: Add the Firefox tests back #263.

🤖 Generated with Claude Code