When Should an Exception Become Part of the Process?

A workflow can handle an unusual case once with a quick workaround. But when the same exception keeps returning, leaving it undocumented means people must improvise; adding every variation to the main route can make the process harder to use.

What tells you an exception deserves its own defined path rather than remaining a case-by-case judgment? Is it the frequency, the consequence of getting it wrong, the effort of the workaround—or something else? And who should own revisiting that decision when the context changes?