And personally I disagree with the notion that an "include" of a
nonexistent file should be treated as a no-op. Mistakes should generate
error reports.
Or if resources permit, incoming messages could be queued for processing
later, once the errors are fixed. This is a gnarly problem, though. The
closest analogy I can think of is a sendmail :include: alias that
references an unreadable file.
This sort of error handling gets very complicated very quickly. It might
be better to just say "deliver to the INBOX" if the include fails.
--lyndon