test(prose): correct a stale PR state and two issue/PR mislabels

[JarryShaw]

Please follow the guide below

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort) — prose-only, no code touched
• make test passes, and a test case covers the change — ran the three touched files individually under pytest and plain unittest (never the full suite); no behaviour changed, so no new coverage is needed
• Added a changelog entry under docs/source/changelog/ and regenerated CHANGELOG.md, if the change is user-visible — N/A, changelog centralised in docs(changelog): shared 1.5.0 changelog — long-lived, merges last (#610, #616, #617, #618, #620) #657

What is the purpose of your pull request?

• fix — corrects a defect
• feat — adds a feature
• perf — changes performance, not behaviour
• refactor — changes neither behaviour nor performance
• test — tests only
• docs — documentation only
• ci — workflows or build tooling
• chore — anything else

---

Description of your pull request and other information

Found while surveying #NNN citations under tests/ for #719. Three factual errors in test prose, unrelated to the citation-style question itself (tests/** is exempt from that rule):

• tests/protocols/application/test_http_unit.py:2361 called PR corekit: let a nested schema's field callbacks reach the enclosing schema #457 "still-open"; it merged on 2026-09-18. The GH-445 citation next to it is correct and left alone — GH-nnn is this repo's own established issue shorthand (used throughout pcapkit/ and tests/), and A nested schema cannot reach the enclosing packet's fields by name, so CGA Parameters raises KeyError: 'length' #445 is in fact an issue.
• tests/test_base_class_contract.py:45 called feat(foundation): make subclass registration opt-in for Engine, Reassembly, TraceFlow and Dumper (#514) #547 and feat(protocols): opt-in code= registration with enum-type inference for Protocol/ProtocolBase #570 "issues"; both are pull requests.
• tests/const/test_const_str_payload_870_unit.py:12 called fix(const,vendor): bring 8 bespoke registries onto EnumRegistry, stop them minting #869 a "GitHub issue"; it is a pull request.

No behaviour change.

[JarryShaw]

GOOD TO GO at adfd21043 — opus cross-review, round 2. Labelled review: good-to-go.

It re-derived every claim in the rewritten paragraph from its own worktree rather than trusting the
author or me. EXPECTED_FAILURES: 43 keys, exactly one httpv2 key (httpv2-frame/PRIORITY), no
key containing SETTINGS at all. The round trip: it loaded the suite's own generator (322 cases, 10
httpv2) and ran each — httpv2-frame/SETTINGS returns status='OK', and PRIORITY is the only httpv2
failure, which is precisely its one recorded entry. The bare pack: KeyError('flags') via
schema.py:866 → httpv2.py:174. It flagged that it doubted this one, since post_process is
documented as revising data after unpacking, so it was not obvious pack calls it — it does.

It settled the dropped-pointer question with evidence I did not have, and the answer is stronger
than "harmless". #457 is merged and #445 is closed, so the original "still-open" was false
rather than stale. More to the point, the residual KeyError is not the defect #445 described.
#445 was a nested schema failing to reach the enclosing packet's fields by name; its fix makes an
undeclared name fall through to the enclosing mapping, and the reviewer confirmed that works —
sf.pack({'flags': {...}}) succeeds, 6 octets. The bare pack fails because there is no parent at
all, so there is nothing to fall through to. A past-tense pointer would therefore imply the
remaining error is a trace of a fixed defect, misleading exactly the reader it was meant to help. So
dropping it was right. #445 is still cited correctly in four other files.

Amend scope confirmed: 1 file, +8/−9, one paragraph; the other two files are byte-identical to round 1.

One finding independent of this PR, which I reproduced myself and am filing separately. Those three
modules pass individually but fail when run together in one process:

tests.test_base_class_contract alone          -> Ran 7 tests ... OK
the three modules in one process              -> Ran 73 tests ... FAILED (failures=1, errors=3)
pytest over the same three files              -> 73 passed, 603 subtests passed, exit 0

RegistrationGateTests.test_user_style_subclass_registers_when_it_opts_in loses
user_opt_in_dumpers. Pre-existing — it reproduces on main — and pytest reports it green, which
is the part worth knowing.