{
  "schema": "anvil-investigation.trace-refill-analysis/v1",
  "status": "diagnostic-only, preflightfailed19/20toolbatch; noqualificationclaim",
  "pairing": "SingleC1 ordered3requests; final3largebatchedlogs map exactly via prompt_tokens=batched_tokens+4windowtokens+1lasttoken. No extra traffic assumed perrootcontrolledrun.",
  "requests": [
    {
      "request_index": 0,
      "prompt_tokens": 7415,
      "batched_tokens": 7410,
      "window_tokens": 4,
      "last_prompt_token_read_in_decode": 1,
      "loan_slots": 1424,
      "loan_chunk_tokens": 7424,
      "lend_ms": 0.2,
      "batched_ms": 10376.4,
      "refill_ms": 13210.8,
      "window_ms": 234.1,
      "trace_components_ms": 23821.5,
      "visible_ttft_ms": 26490.712800001347,
      "first_output_ms": 23981.571899999835,
      "e2e_ms": 26957.813500001066,
      "phase_host_staging_cumulative_ms": 3928
    },
    {
      "request_index": 1,
      "prompt_tokens": 7417,
      "batched_tokens": 7412,
      "window_tokens": 4,
      "last_prompt_token_read_in_decode": 1,
      "loan_slots": 1424,
      "loan_chunk_tokens": 7424,
      "lend_ms": 0.1,
      "batched_ms": 12397.7,
      "refill_ms": 13401.9,
      "window_ms": 164.6,
      "trace_components_ms": 25964.3,
      "visible_ttft_ms": 30198.69739999922,
      "first_output_ms": 26157.44099999938,
      "e2e_ms": 30648.735099999612,
      "phase_host_staging_cumulative_ms": 4649
    },
    {
      "request_index": 2,
      "prompt_tokens": 7418,
      "batched_tokens": 7413,
      "window_tokens": 4,
      "last_prompt_token_read_in_decode": 1,
      "loan_slots": 1424,
      "loan_chunk_tokens": 7424,
      "lend_ms": 0.1,
      "batched_ms": 10818.3,
      "refill_ms": 13972.8,
      "window_ms": 275.8,
      "trace_components_ms": 25067.0,
      "visible_ttft_ms": 28609.902000000147,
      "first_output_ms": 25175.632400001632,
      "e2e_ms": 29049.241700002312,
      "phase_host_staging_cumulative_ms": 5323
    }
  ],
  "cgroup_observations": {
    "before_events_max": 1087870,
    "after_events_max": 2166246,
    "delta_events_max": 1078376,
    "oom_and_oomkill_both_samples": 0,
    "hard_memory_bytes": 53687091200,
    "same_peak_bytes": 53687103488,
    "limitation": "Asynchronous preflight-end/final status boundaries. Delta cannotbeassignedperrequest; memory.maxevents showcharge/reclaimpressure, notOOM orcauseofoneconnectionfailure."
  },
  "conclusion": "Restoring1424borrowedexperts takes13.2–14.0s, exceeding10.4–12.4sbatchedprefill inall3. Materialouterpromptcost confirmed; no claimphysicalSSDlatency. Small fixed4096borrow-retainedtrial justified, samehardlimits, withqualitygateandnoexpectedperformanceguarantee.",
  "source_artifacts": {
    "trace-final-logs.json": "507013edb233656107bbfe3496ad3d2aeb220641493c907018dbecb947ff3c41",
    "trace-final-status.json": "e0f8449cad588ccc875c66d4b94af2492d2f235ade2b1c50f5919e9e7c229081",
    "trace-preflight-failure-status.json": "9957155c02208d0ac4e40e21b12dbfc1155e5064b5db0dc5fd1612f49a8b9f1f",
    "flash-trace-8k-c1.json": "83abeeaa1d945b114c34a43d68e6e69b6b1e4e07d5cb45327f9455242722ce5f"
  }
}