When we stat a filepath, if we get Errno::ENOENT we place the watched file into a delay_delete cycle meaning we need to get two successive Errno::ENOENT errors to actually initiate a cleanup.

We can't really test for these scenarios and so we rely on user feedback to identify them in order to change the code to handle these different errors as if they are equivalent to Errno::ENOENT. Currently any errors that are not Errno::ENOENT are logged at the error log level.

I propose to add a new setting that allows users to provide a set of strings that when converted to regexes are applied to the error message, when matched, will be treated as equivalent to Errno::ENOENT.

Naming is hard and I haven't got a clue as to what name to give to this setting.