In most cases the shortened final -o is from 'iambic' shortening, isn't it? In later texts (eg. Symphosius' riddles) it spreads to ō after a heavy syllable as well, but this is hardly a classical phenomenon. mihi is found very early, and is pretty much regular; at least, 38x in the Eclogues vs. mihī 3x (of course, the former might just be more convenient in hexameter). crēdō is obviously rarer in Vergil, but the two instances in the Georgics and Eclogues both scan - -.