fix(ci): raise AVM check-circuit per-tx timeout to 120s (configurable)

[AztecBot]

Summary

Raise the per-input avm_check_circuit timeout in yarn-project/end-to-end/bootstrap.sh from the hard-coded 30s to a 120s default, configurable via AVM_CHECK_CIRCUIT_TIMEOUT.

Root cause

The avm-check-circuit job on next failed with exit code 124 (timeout): run 28236450261. The job-level exit propagated from a single timed-out child command:

FAILED: bb-avm avm_check_circuit -v --avm-inputs .../e2e_multiple_blobs/avm-circuit-inputs-tx-0x297821c2....bin (32s) (code: 124)

Each check-circuit command runs under ISOLATE=1:TIMEOUT=30s. The e2e_multiple_blobs test emits more than a blob's worth of public data, producing a long AVM trace (~700k rows in a prior occurrence) whose Checking circuit... step needs ~32–35s under the default 2-CPU isolated CI quota — over the 30s budget. There was no circuit assertion or exception; the check was progressing correctly and was killed by timeout at the limit. Every other tx passed, including a second, smaller e2e_multiple_blobs tx (2s). The in-code WARNING comment already anticipated exactly this.

This is not a regression from the triggering commit (#24305, "remove the TS AVM simulator"). That PR only touches the TS simulator package and C++ fuzzer entry points — it modifies neither the C++ AVM circuit checker (barretenberg/cpp/.../vm, the bb-avm binary) nor the check-circuit invocation/timeout in bootstrap.sh/ci.sh. The checker's performance and the timeout are unchanged by it; the heavy tx's check-circuit time has simply crept up to the limit.

History (please merge promptly)

This exact fix has been proposed and lost three times without ever being technically rejected:

• Superseded by #24137: raise AVM check-circuit timeout #23662 (May) → superseded by fix(ci): raise AVM check-circuit per-tx timeout to 120s #24137.
• fix(ci): raise AVM check-circuit per-tx timeout to 120s #24137 ("raise AVM check-circuit per-tx timeout to 120s", same root cause/analysis) was auto-closed as a stale claudebox draft on 2026-06-22 — "Automatically closing this stale claudebox draft PR (no updates for 5+ days). Re-open if still needed." — i.e. it was never merged simply because it sat as a draft.
• A cb/avm-check-circuit-timeout-60s branch from 2026-06-20 carries the same fix with no PR.

The job will keep failing on every heavy run until this lands. This PR reuses #24137's already-reviewed approach (120s default + AVM_CHECK_CIRCUIT_TIMEOUT override) rebased onto current next. Please mark it ready and merge rather than let it go stale again.

Verification

• bash -O extglob -n bootstrap.sh passes.
• Default command generation emits TIMEOUT=120s; AVM_CHECK_CIRCUIT_TIMEOUT=75s emits TIMEOUT=75s.

---

Created by claudebox · group: slackbot