fix(ci): bump avm_check_circuit per-test timeout to 120s for heavy txs

[AztecBot]

Problem

The avm-check-circuit job failed on next (run 27367574268) with exit code 124 — a timeout, not a circuit correctness failure.

The check runs bb-avm avm_check_circuit on every dumped e2e AVM circuit input in parallel, each isolated with TIMEOUT=30s and the default 2 CPUs. Every other input passed in 4–8s; one timed out:

FAILED: .../e2e_multiple_blobs/avm-circuit-inputs-tx-0x11c58d2c....bin (33s) (code: 124)
run_test_cmd '...:ISOLATE=1:TIMEOUT=30s:NAME=avm_cc_e2e_multiple_blobs_0x11c58d2c ...'

Root cause

From the per-test log (--cpus=2):

Generating trace...                          # 18:36:24
Checking circuit...  (mem: 3869 MiB)         # 18:36:42  (~18s for sim + trace gen)
Running check (with skippable) circuit over 700560 rows.
timeout: sending signal TERM to command 'bash'   # 18:36:53  killed mid-check at the 30s cap

This one e2e_multiple_blobs tx produces a ~700k-row circuit — far larger than the other e2e txs (which finish check-circuit in a few seconds), so its trace generation + check exceeds the uniform 30s budget and is killed mid-check. The triggering commit (#24026) is docs-only, so nothing in the circuit changed; the input is simply heavy enough to cross the boundary. TIMEOUT=30s has been unchanged since the feature landed in #18747, and the original WARNING comment in avm_check_circuit_cmds explicitly anticipated this failure mode for larger txs.

This is a recurrence: PR #23771 made the same fix on June 1 but was a draft that got closed unmerged on June 7, so next reverted to the 30s ceiling and the heavy tx times out again.

Fix

Raise the per-test timeout from 30s to 120s in avm_check_circuit_cmds (yarn-project/end-to-end/bootstrap.sh). This gives ~3–4x headroom over the heaviest observed total and room for future circuit growth.

The change is zero-cost for the fast majority of checks — they still exit in a few seconds, so neither parallelism nor stage wall-clock is affected. Only the heavy outlier is allowed to run to completion instead of being killed mid-check. (Deliberately keeping the default 2 CPUs rather than bumping CPUS, which would apply to all ~500 inputs and halve runner concurrency while barely speeding up the already-fast majority.)

No outer workflow/step timeout caps below this value, so the larger per-test budget takes effect.

---

Created by claudebox · group: slackbot

[AztecBot]

Flakey Tests

🤖 says: This CI run detected 1 tests that failed, but were tolerated due to a .test_patterns.yml entry.

\033FLAKED\033 (8;;http://ci.aztec-labs.com/c94d5d4d0d1a9bbe�c94d5d4d0d1a9bbe8;;�):  yarn-project/end-to-end/scripts/run_test.sh simple src/e2e_p2p/sentinel_status_slash.parallel.test.ts "slashes the proposer with INACTIVITY when checkpoint validation records invalid" (204s) (code: 0) group:e2e-p2p-epoch-flakes