Non-technical end users
Same bad address, same returned envelope, but on the Hermes delivery truck this time. The mail system is fair — it rejects both identically.
Campaign a2a-hermes-v0.6.0-r2 FAIL
Infrastructure
Back-filled by scripts/backfill_legacy_runs.sh — historical run predates campaign.meta.json emission.
Run focus
Hermes group hit the same CIDR blocker — identical failure mode, identical fix
What this campaign tested: Hermes campaign dispatched concurrently with openclaw-r2 to confirm the two-VPC concurrent-dispatch model.
What it demonstrated: Workflow concurrency key (group=a2a-gate-${agent_group}) correctly scoped — both campaigns executed in parallel. Both failed identically on the same CIDR, so the blocker is in the shared terraform module, not in anything agent-group-specific.
AI NHI analysis · Claude Opus 4.7
Hermes group hit the same CIDR blocker — identical failure mode, identical fix
RED at Terraform init + apply in 13 s.
For three audiences
C-level decision makers
Confirms cross-campaign fault isolation: a bug in shared infra breaks both groups in the same way, so we see the whole shape of the problem in one dispatch rather than discovering it later with one group already green.
Engineers & architects
Same root cause as r2-openclaw (ip_range 10.260.0.0/24). The DO-side error surfaced at VPC create, not firewall — different from r3-hermes where the VPC succeeded but firewall failed, confirming DO validates each resource in isolation rather than doing a global pre-flight.
What changes going into the next campaign
Dispatch r3 with CIDR fix.