Personally, I'm inclined to call it a bug, as it violates the principle of least surprise - behaving differently based on the order in which subs are defined is a very surprising behaviour. I consider inconsistencies to be a category of bug, so those two options are equivalent, IMO. And I can't imagine how it could possibly qualify as a feature, so "bug" is really the only option left.
However, I consider it to be a bug in the code which reuses the name much more so than a bug in perl itself. It's an excellent example of why one should follow the standard Perl convention of using InitialCaps for module names and all_lowercase for sub names, or at least follow some convention which distinguishes the two types of names. If a programmer chooses to name things ambiguously and gets bitten by that ambiguity, I'd say he deserves the pain that results. Unfortunately, that pain is more often felt by an innocent maintenance programmer rather than by the one who caused the problem in the first place.