Is clearing the CI cache a fix or just a diagnostic step?

A debugging discussion about lockfile drift, package manager installs, and whether cache clearing should count as a root-cause fix.

Clearing CI cache often makes a failed build pass, but I hesitate to call that a fix. If the cache key is wrong, the install command ignores the lockfile, or the package manager version changed, the same failure can return later. I would rather treat cache clearing as a diagnostic step: if it changes the result, document what was stale and adjust cache keys or install commands. Otherwise the team learns only that the pipeline is fragile. How do you record cache-related CI fixes so they do not become recurring folklore?