Mapping peripherial signals to processor pins is something I do directly when I have a favourite processor after having done an initial analysis of the shortlisted alternatives.

It's something that just have to be done before the CAD busines is started up.

The only thing that we can't protect ourselves from is still unpublished errata, or other changes to the published specifications that arrives at a later stage. One time I sent an error report that the two signals D+ and D- that was either GPIO or part of an UART could not have their data direction individually controlled when used as GPIO. The response was that a hew user manual (released about the same day I wrote the error report) added this fact. So the chip vendor decided it wasn't an errata but a fact that was missing from previous versions of the user manual/datasheet. We just can't protect ourselves from such things. But the basic mapping of processor pins should have been done so early in the project that there shouldn't be a need to talk about "switching processor". The processor to use shouldn't have been selected until a valid mapping had been presented.