Any two PRs that add tests conflict on six files by construction (test count hard-coded)

[cdeust]

Problem

The collected test count is hard-coded, by hand, in six files: .bestpractices.json, CLAUDE.md, CONTRIBUTING.md, README.md, docs/ASSURANCE-CASE.md, assets/badge-tests.svg. Each is checked for exact equality against whichever branch's own live pytest --collect-only count runs in CI.

That count is a property of the post-merge tree, not of any one branch: two branches that each add tests compute two different, both-true numbers and must each edit the same six lines to match. Whichever merges second silently overwrites the first's correct figure with its own now-stale one.

Measured on this repo, not hypothesised:

• 6 of 8 open PRs conflicted on exactly these files; two PRs conflicted on nothing else.
• A PR flipped from MERGEABLE/CLEAN to CONFLICTING/DIRTY the instant a sibling PR merged.
• main went red twice between merges purely from this race (PR fix(mutmut): dot the last scripts/ test-loader gap, make pip_constraint_sets.py mutation-testable (#262) #280 synced the number to its own total; fix: drain pending tasks before stopping the AP sync loop (issue #258) #278 then added more tests against a now-stale base).
• The same PR was rebased three times purely to resolve this, and four different mechanical resolvers all produced defects.

Fix (this issue tracks the PR that resolves it)

• assets/badge-tests.svg becomes the ONLY artifact stating an absolute test count.
• Its check moves from exact equality to a monotone floor (committed <= live) in both scripts/check_doc_claims.py (doc_claim_structural.check_badge_floor) and scripts/generate_repo_badges.py (stale_tests_badge) — a committed count that lags the live one is stale-but-true and passes; only an over-claim fails.
• The other five files drop the number entirely, pointing at the badge instead.
• A standing regression test (RepositoryTests.test_no_prose_file_states_the_suite_size_any_more) asserts no scanned file — .bestpractices.json included — states this claim in prose again.
• scripts/check_doc_claims.py (420 lines) and scripts/generate_repo_badges.py (305 lines) were both over the repo's 300-line file cap before this work needed to touch them further; split via Extract Module into doc_claim_sources.py/doc_claim_scan.py/doc_claim_structural.py/repo_badge_catalog.py, verified behavior-preserving by an unchanged test suite before the floor logic was added.

Acceptance test: two throwaway branches off the fix branch, each adding a trivial passing test, merge with zero conflicts in the six files.

[cdeust]

Fixed by #294 (merged). Verified on main @ 3f0522b: none of the five prose/JSON docs carries a per-branch test count any more —

.bestpractices.json
CLAUDE.md
CONTRIBUTING.md
README.md
docs/ASSURANCE-CASE.md

Only assets/badge-tests.svg retains a figure, and it is a generated monotone floor, so adding tests never invalidates it. An ordinary test-adding PR now touches none of the six files, which is what removes the conflict by construction rather than by convention.

This also fixes main flapping red between merges — the failure mode where #280 synced the count to its own total and #278 then added tests against a now-stale base, leaving main red until #283 happened to resync it.