While validating the ContainerNursery HEAD-response framing fix with Kompile CLI 0.0.75, a

Reported (UTC): 2026-07-15 17:05

While validating the ContainerNursery HEAD-response framing fix with Kompile CLI 0.0.75, a complete 332-test run exited 1 with only 'TESTS FAILED (330/332 tests completed successfully, 2 failed)' and did not identify either failing test or print either stack trace. I reran the exact suite with set -o pipefail and durable tee logging to /tmp/cn-head-fix-full-suite.log; that rerun again exited 1, but emitted no TESTS FAILED summary, no failing test name, no AssertionError, and no stack trace. The retained log ends with four generic 'Socket read timeout during communication - test may be running slowly' messages. Impact: the two failing tests cannot be isolated or debugged, violating the required full-stacktrace workflow and forcing expensive 20+ minute full-suite reruns. The four directly affected HTTPS tests all pass individually, and an earlier full run exited 0. Suggested durable fix: TestRunner/CLI must always retain each test identity and its complete failure throwable locally, print them before returning nonzero even if the runner socket closes, and write a machine-readable results file so failures survive terminal truncation or controller disconnects.

Production verification — 2026-07-17

Status: STILL EXISTS. The live ContainerNursery process remains central to the reported mechanism and was consuming 238% CPU with 543 threads; host load average was 60.59 and CPU pressure was near 90%.

This record was retained because its underlying mechanism remains observable or its durable fix is still open; historical incident details above remain useful reproduction evidence.