A post-burst review ritual

[blooop]

Question

0.1.0→0.3.3 shipped in ~36 hours; most of the review's findings are that velocity's residue, and the repo's history shows findings do get converted into mechanisms — after someone looks. Decide a standing practice that keeps the burst fast and schedules the sweep: e.g. a drift/architecture review pass at each release tag or after N merged PRs (a review1st-style self-review, or charting a map like the sibling), versus heavier per-PR gates that tax the burst itself. Name what is checked, when, and by what.

[blooop]

Resolved: the ritual already exists; naming it is the whole deliverable. The practice is: an architecture review chartered as a wayfinder map, run when the reviewer judges enough has changed, never as a per-PR gate. The mechanism that runs every time is the fresh-context adversarial review of each PR by an agent that did not write it, which is where the last three real defects were caught (#427's layering violation, #415's four-hole guard, #428's dropped Absent). Nothing new is added; a scheduled trigger is refused because the charter month produced three reviews unprompted, so the trigger was never the missing part. Recorded in CLAUDE.md beside the #319 rule (same PR).