OK, it's not a bug (i.e., a design error) per se, but unless the data sheet rather prominently calls out the "unexpected" behavior, I also wouldn't call it a very robust design. I.e., if you had a piece of software that computed the difference between two numbers, even if the instructions for the software told you that the mean of those numbers couldn't be more than a certain value, few users would think it reasonable if the software actually flipped the sign of the result on them just because they went from asking for 10-5 to 100-95, you know? (This is what some op-amps do -- the common-mode voltage higher than the data sheet's allowed spec and they hit the opposite rail...)
(I would accept that if your common mode voltage on an op-amp exceeds the power supply rails themselves, one can't reasonably expect any particular output behavior -- the part might as well burst into flames :-) -- unless the data sheet explicitly claims otherwise.)
I once worked with a consultant (he did database programming) who charged something like half his regular rate for fixing bugs -- I thought that was at least a nice nod towards wanting to actively encourage himself not to introduce bugs in the first place.
---Joel