fix(qso-answerer): D-CALLER-020 — partner re-transmitting own CQ no longer false-aborts QSO

[frank001]

Summary

• QsoAnswererService.HandleWaitReportAsync and QsoCallerService.HandleWaitRr73Async aborted the QSO the instant they decoded any message from the partner not addressed to us — including the partner simply re-transmitting their own CQ (dest == "CQ") because they hadn't decoded us yet, which is normal early in a retry sequence, not evidence they've moved on.
• Caught in real-hardware field testing (logs/openswfz-20260714T171230Z.log, EA1FUB): one retry fired correctly, then the partner's own CQ re-decode ended the QSO outright ("timing was impeccable, but the application just decided to end the whole qso").
• Fix: exclude dest.Equals("CQ", OrdinalIgnoreCase) from the abort condition in both services; falls through to the existing "no matching message" retry/watchdog backstop instead. Genuine third-party addressing (a different real callsign) still aborts immediately, unchanged.
• Pre-existing defect, not introduced by PR fix(qso-answerer): D-CALLER-018/016/019 — abort hard-stop, faster re-engage, dblclick cleanup #72 — confirmed present on main before that merge.

Test plan

• New unit test in QsoAnswererServiceTests: partner re-transmitting own CQ in WaitReport does not abort; drives the same retry-exhaustion cycle pattern as the existing silence-based test to confirm the retry backstop still eventually gives up (AC-1, AC-5).
• New unit test in QsoCallerServiceTests: same for WaitRr73 (AC-3).
• Existing 6.6 (WaitReport_PartnerWorksOther_AbortsToIdle) and 5.9 (WaitRr73_PartnerWorkingOther_Aborts) genuine-third-party tests pass unmodified (AC-2, AC-4).
• Existing WaitReport_NoResponse_RetriesThenAborts (line 460) retry-exhaustion test passes unmodified (AC-5).
• dotnet build OpenWSFZ.slnx -c Release — 0 warnings, 0 errors (AC-6).
• Full suite green, no regressions (AC-7).
• Live verification (dev-task §5): added qa/d-caller-020-working-cq-false-abort-verify/, a standalone tool that genuinely synthesises + decodes real FT8 audio (qa/rr-study/synth_wav.py + the real Ft8Decoder/libft8 P/Invoke) and drives the real, unmodified QsoAnswererService/QsoCallerService with it — stronger than unit tests, which never round-trip through real FT8 packing. Verified PASS with the fix; a deliberate control run with the fix reverted reproduces the exact regression on both services (see live-reports/).

🤖 Generated with Claude Code

[frank001]

QA sign-off. Reviewed the actual diff, not just the commit message.

Fix matches the dev-task exactly — narrows the abort condition in both QsoAnswererService.HandleWaitReportAsync and QsoCallerService.HandleWaitRr73Async to exclude dest == "CQ", falling through to the existing retry/watchdog backstop instead of a same-cycle snap abort. Minimal, symmetric, no scope creep.

Tests: new coverage drives the full retry-exhaustion cycle proving both halves — partner's own CQ no longer aborts, and the existing retry-count backstop still eventually gives up on genuine silence (no infinite-retry regression introduced). Existing 6.6/5.9 genuine-third-party tests pass unmodified, confirming the correct-abort case is untouched.

Live verification exceeds the dev-task's own bar — the harness re-synthesises and re-decodes real FT8 audio through the actual Ft8Decoder/libft8 P/Invoke rather than replaying a log or hand-authoring DecodeResult strings, then drives the real production service classes. Includes a genuine control run (fix reverted → FAIL on both services, reproducing the exact field regression; fix reapplied → PASS) — a real regression guard, not a rubber stamp.

CI green on all three platforms, mergeable. Approving to merge as-is.