F335 atomic completed-epoch retirement checks Date: 2026-08-07 UTC Host: cuda-ke Kernel: Linux 7.0.0-28-generic x86_64 Python: 3.12.3 MPI: Open MPI 4.1.6 / mpi4py 4.1.1 / MPI 3.1 Static and syntax checks - ruff touched production/test/evidence Python: pass - Python bytecode compilation for touched Python: pass - git diff --check including untracked delivery files: pass - evidence-directory __pycache__ residue: none Tests - distributed/lifecycle/filesystem directed tests: 195 passed, 30 subtests in 15.74 s - related MPI/distributed/hybrid tests: 312 passed, 38 subtests in 16.62 s - full warnings-as-errors Python regression: 708 passed, 58 subtests in 96.23 s - retirement/GC negatives: active symlink, malformed reserved name, retired symlink, invalid limit, and post-rename fsync uncertainty all fail closed Real MPI post-rename uncertainty and convergence - ranks and masters/workers: 7 and 2/5 - sealed work epoch: f335 repeated 16 times - actual MPI transport: true - actual multi-host: false - synthetic topology: false - injected failure point: rank-0 rename visible, parent-directory fsync not performed - injected worker ACKs before retirement failure: 3/3 + 2/2 - injected return code and elapsed: 72 and 3.6238950230181217 s - active after injected failure: false - exact retired roots after injected failure: 1 - matching recovery startup reclaimed roots: 1 - matching recovery worker ACKs: 3/3 + 2/2 - matching recovery return code and elapsed: 0 and 3.7320157941430807 s - active after matching recovery: false - exact new retired roots after matching recovery: 1 - prior and new retirement identities differ: true - content-addressed seed digest and bytes intact: true - all JSON checks: pass Mechanism cost - warm-ups/repetitions per scale: 5/30 interleaved and alternating - 3 files recursive/retirement median and p95: 2593.562/2721.449 vs 2620.398/2661.56 us - 3 files median saved and speedup: -26.83600000000024 us and 0.9897588076315124x - 129 files recursive/retirement median and p95: 3280.4790000000003/3865.488 vs 2638.7405/2732.876 us - 129 files median saved and speedup: 641.7385000000004 us and 1.2431987912415035x - 1025 files recursive/retirement median and p95: 9008.330999999998/10274.199 vs 2875.4610000000002/3827.281 us - 1025 files median saved and speedup: 6132.869999999998 us and 3.1328301792303903x - raw sample counts: 30+30 at each of three scales - all samples complete: true Strict evidence boundaries - deployment_evidence=false - remote_power_loss_tested=false - real_fsync_failure_tested=false - actual_multi_host=false - cross_job_lock_exclusion_tested=false - nfs_server_failover_tested=false - mpi_rank_repair_tested=false - network_partition_tested=false - ulfm_repair_tested=false - end_to_end_campaign_benchmarked=false - dse_throughput_benchmarked=false - solver_or_coverage_uplift_measured=false - lava_m_campaign_run=false - GC limit bounds retired roots, not entries or wall time Repository inventory after F335 - documented feature IDs: F00-F335 (336 total) - documented SYMCC configuration names: 411 - top-level test files: 246