diff --git a/yarn-project/end-to-end/bootstrap.sh b/yarn-project/end-to-end/bootstrap.sh
index c08a2f2374d4..9c8766040d1c 100755
--- a/yarn-project/end-to-end/bootstrap.sh
+++ b/yarn-project/end-to-end/bootstrap.sh
@@ -198,11 +198,14 @@ function avm_check_circuit_cmds {
   # Commands run from repo root via parallelize, so use path from top
   local dump_dir_from_top="yarn-project/end-to-end/$default_avm_inputs_dump_dir"
 
-  # Specify timeout and resources
-  # WARNING: theoretically, transactions could need more CPU and MEM than we allocate by default.
-  # In that case, they might start timing out. For now, all of the e2e test txs seem to be relatively
-  # small and the AVM can run check-circuit with limited resources.
-  local prefix="$hash:ISOLATE=1:TIMEOUT=30s"
+  # Specify timeout and resources.
+  # Most e2e txs produce small circuits that check in a couple of seconds on the default 2 CPUs, but
+  # larger ones (e.g. an e2e_multiple_blobs tx yielding a ~700k-row circuit) do not finish the check
+  # within 30s when capped at 2 CPUs, so they time out (exit 124). check_circuit parallelises over
+  # rows, so we give each container 4 CPUs to roughly halve the check time plus a larger timeout for
+  # headroom. With `parallelize 16` this is 16*4 = 64 CPUs, matching the smallest CI fleet instance
+  # (no oversubscription).
+  local prefix="$hash:ISOLATE=1:TIMEOUT=90s:CPUS=4"
 
   # Find all .bin files in the dump directory (handles nested dirs)
   for input_file in "$default_avm_inputs_dump_dir"/*/*.bin "$default_avm_inputs_dump_dir"/*/*/*.bin; do