Say How Each Measured Requirement Actually Fails - #648

The rationale collapsed two requirements into one failure mode and got it wrong for
one of them. It said an older interpreter fails "as an AttributeError at the call
rather than at startup", which is true of str.removeprefix and false of
datetime.UTC: that arrives through a module-level `from datetime import UTC` in
spec/audit.py, scripts/pr_review.py and its tests, so an older interpreter raises
ImportError before any of the module runs.

Both failures were constructed rather than described, an ImportError from a
module-level import of a name that does not exist, and an AttributeError from a
call on a str without the method. The entry now names each requirement with its own
mode and says the import failure is the one a host meets first, since the script
does not start.

This text is what host_gate.py prints to a host that fails the floor, so the mode
it names is the symptom that reader is looking at.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

"The second is the one a host meets first" was a blanket claim over a per-module
fact. scripts/prose_lint.py calls removeprefix and imports no UTC, so a host running
that script sees the AttributeError path and never reaches an import failure.

The three modules divide cleanly, checked against the tree rather than recalled:
spec/audit.py carries both requirements and fails at import, scripts/pr_review.py
carries only the import and fails the same way, and scripts/prose_lint.py carries
only the call and so starts, runs and fails partway through. The entry now says the
mode is decided by the script a host runs rather than by the interpreter alone.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Two corrections, one from the review and one from auditing the rest of the field
after it, which is the pass that should have happened four findings ago.

The named one: the entry said both removeprefix and removesuffix are called in
spec/audit.py and scripts/prose_lint.py. Enumerating call sites per method rather
than with one combined grep gives removeprefix at prose_lint.py:91 and audit.py:714,
and removesuffix at audit.py:74 and audit.py:715 only. The combined grep is what
produced the combined claim.

The one the audit found: the entry said CI gates "the standard-library script tests
and the prose and repo gates and nothing else". It also runs markdownlint, cspell,
actionlint, editorconfig-checker and spec/validate.py. The point being made, that no
Python linter or type checker runs there, survives, and the "nothing else" did not.

Every remaining claim was checked against the tree in the same pass. Standard
library only holds, verified by walking every import in every tracked .py against
sys.stdlib_module_names and the local modules, with no third-party import found. The
pyproject settings, the UTC import sites and the per-module failure modes each read
as stated.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>