When a failing dependency should stop being called

The pattern is simple enough to explain in a sentence, after enough consecutive failures, stop calling the thing and fail fast until it looks healthy again, and the reason it matters is less obvious.

Without it, a slow dependency does not cause a slow feature. It causes a queue. Every request waiting on a five-second timeout is a worker not serving anyone else, and on a shared host that is how one third-party outage becomes your outage. The failure that takes a site down is rarely the dependency dying; it is the dependency getting slow while everything keeps politely waiting for it.

The monitoring angle in the piece is the part worth stealing. A breaker that has tripped is a signal you can alert on, and it is a far better signal than error-rate alone because it tells you the system already knows something is wrong.

For WordPress work this is directly relevant and widely ignored: licence checks and update pings on admin page loads are exactly this shape, and a breaker plus a cached last-known-good answer is usually a small change with a large effect.