PR #33402 real dev verification
main: d1aeed249288585390dbac4fde5b54c35bd5fc06
PR head: c54e745f69548f5da62169e0f4deae24fc012a88
client: repo Playwright 1.60.0 + /usr/bin/google-chrome
provider: Codex gpt-6.1-sol through local proxy

Created all three sessions through the web composer in /home/user/import-history-check.
Stopped the generating server, started fresh isolated main state, and clicked Welcome > Import 1 project and start.
Stopped main before starting the PR with fresh isolated state and repeated the same UI import.

$ node /tmp/import-report.mjs <isolated-state-db> <stage>
(read-only query of running server projections; summary follows)
The native transcript also contains one ambiguous setup-context record retained by both revisions, so full import has 211 records for the 210-message conversation.
{
  "stage": "source",
  "threads": [
    {
      "title": "long",
      "messages": 210,
      "users": 105,
      "assistants": 105,
      "historyAcks": 105,
      "missingAcks": [],
      "duplicateAcks": [],
      "notices": []
    },
    {
      "title": "short",
      "messages": 2,
      "users": 1,
      "assistants": 1,
      "historyAcks": 0,
      "notices": []
    },
    {
      "title": "repeat",
      "messages": 4,
      "users": 2,
      "assistants": 2,
      "historyAcks": 0,
      "notices": []
    }
  ]
}
{
  "stage": "main",
  "threads": [
    {
      "title": "repeat",
      "messages": 5,
      "users": 3,
      "assistants": 2,
      "historyAcks": 0,
      "notices": []
    },
    {
      "title": "short",
      "messages": 3,
      "users": 2,
      "assistants": 1,
      "historyAcks": 0,
      "notices": []
    },
    {
      "title": "long",
      "messages": 200,
      "users": 100,
      "assistants": 100,
      "historyAcks": 100,
      "missingAcks": [
        "HISTORY_ACK_001",
        "HISTORY_ACK_002",
        "HISTORY_ACK_003",
        "HISTORY_ACK_004",
        "HISTORY_ACK_005"
      ],
      "duplicateAcks": [],
      "notices": [
        "11 earlier messages were not imported. This thread keeps the first user message and the most recent history. The original session transcript is unchanged."
      ]
    }
  ]
}
{
  "stage": "pr",
  "threads": [
    {
      "title": "repeat",
      "messages": 5,
      "users": 3,
      "assistants": 2,
      "historyAcks": 0,
      "notices": []
    },
    {
      "title": "short",
      "messages": 3,
      "users": 2,
      "assistants": 1,
      "historyAcks": 0,
      "notices": []
    },
    {
      "title": "long",
      "messages": 211,
      "users": 106,
      "assistants": 105,
      "historyAcks": 105,
      "missingAcks": [],
      "duplicateAcks": [],
      "notices": []
    }
  ]
}
{
  "stage": "after-reimport",
  "threads": [
    {
      "title": "repeat",
      "messages": 5,
      "users": 3,
      "assistants": 2,
      "historyAcks": 0,
      "notices": []
    },
    {
      "title": "short",
      "messages": 5,
      "users": 3,
      "assistants": 2,
      "historyAcks": 0,
      "notices": []
    },
    {
      "title": "long",
      "messages": 213,
      "users": 107,
      "assistants": 106,
      "historyAcks": 105,
      "missingAcks": [],
      "duplicateAcks": [],
      "notices": []
    }
  ]
}

UI checks: earliest HISTORY_ACK_001 is visible after paging to the start on PR; main starts at HISTORY_ACK_006.
Resumed long imported session: asked for its original codeword without supplying it; Codex replied amber-orchid-33402.
Resumed short imported session: Codex replied violet-maple-33402.
Repeat session: DOM contained exactly 2 identical user prompts and 2 REPEAT_IMPORT_ACK replies.
Re-import after continuation: 3 imported threads; message counts unchanged (long 213, short 5, repeat 5).

$ blacksmith-check run --id <task-owned-id> "vp i && vp test run apps/server/src/project/AgentSessionScanner.test.ts apps/server/src/project/AgentSessionImporter.test.ts"
Test Files 2 passed (2)
Tests 65 passed (65)
Scanner 61 tests; Importer 4 tests.

No source changes, commits, or pushes were needed.
Not exercised: real Claude/Pi sessions, native desktop/mobile clients, relay/tunnel connections, huge transcript read-budget boundaries.