#1973 fix(broker): log the whole error chain when a worker fails to spawn

Why

On fleet node sf-frame, claude spawns fail with only failed to spawn worker about 5ms after start (crates/broker/src/worker.rs:1670, command.spawn().context("failed to spawn worker")). Example from sf-frame.log.2026-10-12:

2026-10-12T21:41:45.003167Z ERROR relay_broker::runtime::relaycast_events: failed to spawn worker via relaycast WS child=e2e-d1fix-sfframe-1 error=failed to spawn worker

relaycast_events.rs logged the anyhow::Error with Display (%e), which prints only the outermost context. The OS error that explains the failure (for example No such file or directory) was dropped. The fleet action.result already uses {error:#}, but on that node those results are rejected with d1_pressure, so the log line is the only place the cause would appear.

Change

The log line and the eprintln! now use {:#}, through a spawn_failure_detail helper. This changes log output only. Behaviour is unchanged.

Verification

$ cargo test -p agent-relay-broker --lib spawn_failure_detail
test runtime::relaycast_events::tests::spawn_failure_detail_keeps_the_os_error ... ok
test result: ok. 1 passed; 0 failed; ...

Mutation check. I changed the helper to format!("{error}"):

test runtime::relaycast_events::tests::spawn_failure_detail_keeps_the_os_error ... FAILED
test result: FAILED. 0 passed; 1 failed; ...

After restoring it (cmp identical), the test passes again. cargo fmt --check and cargo clippy -p agent-relay-broker --lib -- -D warnings are clean.

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

---

[!NOTE]
Low Risk
Logging-only change in one error path; no spawn, dispatch, or control-plane behavior is modified.

Overview
Relaycast WebSocket worker spawn failures now log the full anyhow error chain instead of only the outer context string failed to spawn worker.

The tracing::error! line and matching eprintln! in relaycast_events.rs go through a small spawn_failure_detail helper that formats with {:#}, so operators see underlying causes (for example a missing binary / NotFound) when fleet action results are unavailable. Spawn behavior and error propagation are unchanged.

CHANGELOG documents the fix under Fixed.

^{Reviewed by Cursor Bugbot for commit aca562f7d12e8bc7f87e98c0250868c1bb43f53f. Bugbot is set up for automated code reviews on this repo. Configure here.}

---

Agent Relay sessions

• claude session a8232a06 · opened via gh pr create · last active 2026-10-13