The CI restructure moves most jobs into a single main.yaml dependency graph wired with needs, minimising separately-triggered workflows. benchmark-seeding.yml was never covered by that plan and needs a decision.
It measures seeding throughput against a swarm of leecher pods — a benchmark, not a correctness gate, so it does not obviously belong on every push.
Options:
- fold it into
release.yaml, so a tagged build records a throughput number;
- keep it separately dispatchable, accepting one more trigger against the minimise-triggers goal;
- delete it, and re-add a benchmark when there is a reason to track the number over time.
Deferred deliberately — not blocking the restructure, and the restructure should not silently decide it either.
https://claude.ai/code/session_013dEiCgEMNhNAy1Rn7Mm2KL
The CI restructure moves most jobs into a single
main.yamldependency graph wired withneeds, minimising separately-triggered workflows.benchmark-seeding.ymlwas never covered by that plan and needs a decision.It measures seeding throughput against a swarm of leecher pods — a benchmark, not a correctness gate, so it does not obviously belong on every push.
Options:
release.yaml, so a tagged build records a throughput number;Deferred deliberately — not blocking the restructure, and the restructure should not silently decide it either.
https://claude.ai/code/session_013dEiCgEMNhNAy1Rn7Mm2KL