Now, some algorithms and routines may be more robust and allowing than others. Maybe, an innocent refactoring attempt just lost that critical bit of robustness, required to deal with that particular format produced by this particular application.
For example, consider an XML-based format, where a particular application delivered a malformed document, like a missing closing quote for some attribute. Most XML interpreters will churn happily along with this, but, after a rewrite of some routine, an application just ignores the malformed tag with the runaway string. Did it break XML? Or did it just fail to interpret a malformed document, it had somehow been able to deal with thanks to some extra robustness present in the previous version?
Considering this hypothetical case: Should that application be improved by an update to regain its previous robustness? Yes, absolutely. Is it a bug and is the vendor to blame? Probably not. Mind that this might be quite well what is happening here, as well.