fix(ci): give e2e_multiple_blobs avm_check_circuit more CPUs and time

[AztecBot]

What

Bump the per-test resource budget for avm_check_circuit on e2e_multiple_blobs transactions from the default TIMEOUT=30s / CPUS=2 to CPUS=8:TIMEOUT=180s in yarn-project/end-to-end/bootstrap.sh.

Why

The avm-check-circuit job has been failing on every next commit (runs 1786 Jun 14, 1787, 1788, and 1789 — the run that triggered this fix). All of them fail on the same test: bb-avm avm_check_circuit for an e2e_multiple_blobs tx, exiting with code 124 (the timeout -v 30s wrapper killing it).

Per-test log for the failing run (ci.aztec-labs.com/ca2b62a6682db814):

Resources: CPUS=2 MEM=8g TIMEOUT=30s
09:50:17 Simulating...
09:50:18 Generating trace... (mem: 825.80 MiB)
09:50:36 Checking circuit... (mem: 3911.55 MiB)
09:50:36 Running check (with skippable) circuit over 700560 rows.
09:50:47 timeout: sending signal TERM to command 'bash'

The e2e_multiple_blobs tx produces a ~700k-row AVM circuit. With only 2 CPUs, trace generation alone takes ~18s and the circuit check then runs out of the remaining ~11s before the 30s wall-clock timeout fires. Every other dumped tx checks in 4–7s, so the tight default budget is correct for them — only the multi-blob circuits are large enough to time out.

This is exactly the scenario the WARNING comment in avm_check_circuit_cmds anticipated ("transactions could need more CPU and MEM than we allocate by default … they might start timing out").

Change

• Add a per-test resource override (same case pattern already used in test_cmds): e2e_multiple_blobs* gets CPUS=8:TIMEOUT=180s; all other txs keep the tight 30s/2-CPU default to preserve parallelism. More CPUs cut both trace generation and the row check (both parallel), and the higher timeout leaves generous headroom as the circuit grows.
• Reword the WARNING comment to point future maintainers at the new case statement when another test dir starts timing out.

Verified the command generator emits ...:CPUS=8:TIMEOUT=180s:NAME=avm_cc_e2e_multiple_blobs_... only for the multi-blob inputs and leaves the others untouched (the later TIMEOUT overrides the default per ci3/source_test_params). MEM defaults to CPUS*4 = 32g, well above the observed 3.9 GiB peak.

This is a CI resource/timeout fix only — no AVM circuit or simulation logic changes.

---

Created by claudebox · group: slackbot