[V3RON]

Closes #759

Status

Implement: done (4/4, audit probed 4/4) Review: round 1 + fix-check passed Hardware: n/a Gate: waiting for the stack
Next: gate the stack (#755 → #775 → #783 → #784 → #785 → #786)

Done when

• pnpm test:changed passes.
• rg -n "workers\.json" src docs --glob '!docs/internal/adr/**' finds only the gateway's old-file list, its test and the docs/CLI.md upgrade section.

Feature completion (#754)

• A worker restarted with fake drivers keeps devices, leases, requests, components and tokens: restart tests in feat(state): the worker's registry lives in state.db #775 (src/daemon/main.test.ts), real tables in feat(drivers): iOS and Android device-data tables #783, tokens in feat(auth): tokens live in state.db on both daemon kinds #784.
• A gateway restarted keeps tokens and drains: feat(auth): tokens live in state.db on both daemon kinds #784 (join token across restart) and the e2e restart test here.
• A renew changes one row with 1,000 requests stored: test in feat(state): the worker's registry lives in state.db #775.
• Refusals for each old file, a damaged or newer state.db, printed by simlock status at once with exit 1: feat(state): the worker's registry lives in state.db #775, feat(auth): tokens live in state.db on both daemon kinds #784, this PR, and feat(cli): the CLI and MCP show why an auto-started daemon did not start #785's e2e (e2e/start-refusal.test.ts).
• docs/CLI.md upgrade and downgrade steps, no live JSON file named: feat(state): the worker's registry lives in state.db #775's section plus each PR's rg Done when line.

Assumptions

none

Review

Spec review: 0 blocking, 0 fixed, 0 notes. Code review: 2 blocking, 2 fixed, 0 notes. Claims review: 0 blocking, 0 fixed, 0 notes.
Rounds: 1. Tokens: 20.4M.
Runs: 9. Cost USD: 2.3576. Agent seconds: 1557.
Assessment: complete

Written by an agent.

[V3RON]

Review ledger

Runs: 9. Cost USD: 2.3576. Agent seconds: 1557.
Tokens input 276, output 49905, cache read 5133982, cache write 346287.
- build implementer: claude-sonnet-5-5 497 s $0.4442
- audit implementer: claude-sonnet-5-5 147 s $0.1753
- review 1 reviewer: claude-sonnet-5-5 357 s $0.1787
- review 1 spec: claude-sonnet-5-5 32 s $0.1327
- review 1 code both: claude-opus-5-5 242 s $0.9575
- review 1 claims: claude-sonnet-5-5 59 s $0.142
- fix 1 implementer: claude-sonnet-5-5 121 s $0.1273
- review fix-check reviewer: claude-sonnet-5-5 35 s $0.0685
- review fix-check reviewer: claude-sonnet-5-5 67 s $0.1314
Finding outcomes: {"fixed":2}. Classes: {"testing 7":1,"testing 8":1}.

{
 "findings": [
  {
   "id": "R1",
   "role": "code",
   "class": "testing 7",
   "outcome": "fixed",
   "path": "src/daemon/main.test.ts",
   "fix": "610f3dee"
  },
  {
   "id": "R2",
   "role": "code",
   "class": "testing 8",
   "outcome": "fixed",
   "path": "src/gateway/drain-store.test.ts",
   "fix": "610f3dee"
  }
 ]
}```

_Written by an agent._