[JarryShaw]

• Searched for similar pull requests
• Followed the coding style (make pylint, make mypy, make isort)
• make test passes, and a test case covers the change
• Added a changelog entry — N/A — 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?

• perf — performance improvements

---

Description of your pull request and other information

Closes #2113. An extraction with engine='pypcap' or 'pcap_ct' probed the pcap backend three times: _select_engine called unsupported_reason() in both the constructor and run(), and the engine constructor called _pcap_backend.probe() again. The default engine never probes.

• Fix: _pcap_backend.Memo is a context manager backed by a ContextVar, so threads don't share it. While one is entered, probe() reuses its result. Extractor keeps one memo per extraction (_exprb) and enters it around both unsupported_reason() calls and the engine construction.
• Not a process-wide cache: probe() is documented as uncached, and many tests mock the distribution scan, sys.modules or import_module and expect a fresh answer on each call. Outside a memo nothing changes. Engine choice and warning text and stacklevel stay the same.
• One behaviour change: with auto=False, run() reuses the constructor's probe, which matches the format decision that probe already made.
• Measured (3.14, stub pcap_ct module, real distribution scan): 0.797 → 0.343 ms per extraction, with scans going from 3 to 1. The default engine is unchanged.
• Test: test_pcap_backend_probe_once_2113_unit asserts one scan per extraction for each pcap engine, the collision warning firing once, and no caching across extractions or across direct calls. Without the fix it fails with 3 != 1. The foundation ordering leg runs 722 tests, 0 failures.

[JarryShaw]

GOOD TO GO at 5e5737db5 (Sonnet review).

• Memo unwinding: an exception inside a nested with memo: with memo: unwinds the token stack, and _MEMO returns to None.
• No stale probe across extractions: the memo lives on the Extractor. __iter__ and __next__ never probe. _exprb always exists by the time run() enters it.
• Behaviour unchanged: the probe-once test fails with the change reverted. The backend, pypcap, pcap_ct, pypcapfile and test_core tests pass.
• Known limit, no change: one Extractor shared across threads could exit its memo out of order.
• UNVERIFIED: the Memo docs entry was read, not rendered.

[JarryShaw]

Ready to merge at 5e5737db5.

• Review: the Sonnet review is GOOD TO GO.
• CI: every blocking check passed (Required checks passed SUCCESS). The 3.8–3.10 KTLO legs carry no warnings.
• Ordering: the tests/foundation plain-unittest ordering leg passed on a scratch merge into main (82b901b91). It ran 722 tests with 0 failures.

This closes #2113.

[github-actions]

Coverage: 91.54% (unit tier, Python 3.15, 5e5737db5, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	4	20	2	95.16%
pcapkit/const	19599	880	2618	866	92.12%
pcapkit/corekit	2580	75	814	37	96.46%
pcapkit/dumpkit	282	3	102	3	98.44%
pcapkit/foundation	2968	23	1034	25	98.80%
pcapkit/interface	138	4	52	3	96.32%
pcapkit/protocols	17414	219	4574	185	98.14%
pcapkit/toolkit	869	36	298	10	95.20%
pcapkit/utilities	473	7	136	5	98.03%
pcapkit/vendor	4432	1988	1018	167	51.87%

Per-file detail: the coverage-html artifact of this run.