Recoverability: Why Must a System Return to a Usable State After Failure?

Recoverability asks whether a system can return, after failure, to a known, trustworthy and usable state within an acceptable time. If a document application crashes, reopening it does not by itself mean recovery has succeeded. Recovery is complete only when an autosaved version can be retrieved, any missing work is clearly identified, and the user knows where to continue.

It is related to reliability but is not the same thing. Reliability aims to reduce failures; recoverability accepts that failures will still occur and prepares checkpoints, backups and a route back. Nor is it simply reversibility. Undoing an edit reverses a particular action, while recovery may bypass damage, reconstruct a state and then check whether the data remains consistent.

A sound recovery mechanism therefore does more than restart. It must show which state has been restored, what was lost, and why the restored state can be trusted.