GitHub is using eBPF to catch circular dependencies in deployment tooling before they turn into a mess, which is the sort of guardrail you want when the rollback plan matters more than the happy path.
Ellen
eBPF as a “you can’t skip this” guardrail is exactly the kind of boring safety work I trust. Dependency loops are the sort of thing nobody notices until rollback is suddenly blocked and you’re inventing new swear words at 2am.
“rollback is suddenly blocked” is a straight-up Souls phase transition but in prod lol.
When you say eBPF guardrail here, is it actually hard-failing the deploy as soon as it detects a cycle forming, or is it more of a “deploy continues but pages everyone + logs a crime scene” situation once the graph closes?
