DCC/protocol stress verification - 2026-10-27

Production DCC/protocol sources unchanged.
Final test source: tests/test_dcc_stress.c
Deterministic seed: 2587695045
Each run: 200000 parser/sanitizer/split mutations and 80000 offer lifecycles,
then bounded local-loopback transfer cycles. No public IRC/network testing.

Normal sustained run (not repeated):
  Requested live time: 90 seconds
  Total elapsed: 90.972106 seconds
  Completed transfer/offer lifecycles: 119952
  Live cycles: 4439
  Verified payload bytes: 1126498344
  Initial/final descriptors: 3 / 3
  Exit: 0
  The original log's rough transitions counter is superseded by the exact
  lifecycle formula: 80000 + 9 * live_cycles + 1.

Final ASan/UBSan + LeakSanitizer run:
  ASAN_OPTIONS=detect_leaks=1
  UBSAN_OPTIONS=halt_on_error=1
  Requested live time: 60 seconds
  Total elapsed: 62.295490 seconds (wrapper: 62.346145 seconds)
  Completed transfer/offer lifecycles: 109359
  Live cycles: 3262
  Verified payload bytes: 827782624
  Engine calls: 165277
  Initial/final descriptors: 3 / 3
  Peak observed descriptors: 20 (fixture peers included)
  Exit: 0; no ASan, UBSan or LeakSanitizer diagnostic

The first leak-enabled attempt had an incomplete log despite a resumed tool
exit of zero. That log is retained and not counted as a successful run.
The final run records the actual child exit in a separate JSON file.

Scope: active SEND engine, parser and temporary-file lifecycle testing on
Linux; timeouts are advanced through the supplied monotonic-time API.
These measurements include fixture overhead and are not full-client
benchmarks or evidence of indefinite uptime. No production defect was
reproduced in this DCC/protocol workload.
