Every check must be able to fail

A green suite and a suite that cannot go red look identical

Bothy has a CI job whose only purpose is to break something on purpose — because a check that has never failed might not be a check at all.

Every test suite is green until it isn't. The trouble is that a suite which passes and a suite which cannot fail look exactly the same from the outside — same tick, same colour, same line in the pull request.

So Bothy has a CI job whose only purpose is to break something on purpose.

It plants a mutation in the code a check is supposed to catch, runs that check, and fails the build if the check stays green. The check has to prove it can go red before its green is worth anything.

Why this is not paranoia

The checks most likely to rot are the ones nobody looks at, because they have never failed. A grep that silently stopped matching after a rename. An assertion guarded by a condition that is now always false. A test whose fixture drifted until it asserts nothing at all.

None of those announce themselves. They just quietly stop being checks, and the build stays green the whole way.

What it costs

One job, a few seconds. Against that: certainty that the thing guarding your release can still say no.