What happens
Running several redcode acp agents in parallel on one machine (redcode 0.8.1, spawned as child agents by the redskilled daemon's Workers), one of them died mid-turn with:
level=ERROR run=b9e20805 message=process session.id=ses_fe04d3246ffekrY9MBvupXLmO1 error="Failed to execute statement" stack="Error\n at <anonymous> (/$bunfs/root/chunk-…"
All instances share ~/.red/redcode/data/opencode.db (plus its WAL) and one data/log/opencode.log. Under 3–5 concurrent agents the statement failure killed the whole turn; the supervising Worker saw an internal error.
Expected
Parallel headless redcode acp processes are a first-class scenario (one machine runs N autonomous agents). Either the shared store handles concurrent writers (busy_timeout/retry on SQLITE_BUSY), or each ACP session can be pointed at an isolated data dir (env override), so one agent's write never aborts another's turn.
Environment
redcode 0.8.1 (npm, bun-compiled), Linux WSL2, 3–5 simultaneous redcode acp children, 2026-08-20.
What happens
Running several
redcode acpagents in parallel on one machine (redcode 0.8.1, spawned as child agents by the redskilled daemon's Workers), one of them died mid-turn with:All instances share
~/.red/redcode/data/opencode.db(plus its WAL) and onedata/log/opencode.log. Under 3–5 concurrent agents the statement failure killed the whole turn; the supervising Worker saw an internal error.Expected
Parallel headless
redcode acpprocesses are a first-class scenario (one machine runs N autonomous agents). Either the shared store handles concurrent writers (busy_timeout/retry on SQLITE_BUSY), or each ACP session can be pointed at an isolated data dir (env override), so one agent's write never aborts another's turn.Environment
redcode 0.8.1 (npm, bun-compiled), Linux WSL2, 3–5 simultaneous
redcode acpchildren, 2026-08-20.