My impression here is that most of the frustration is due to people not using any HiDPI/Qt5 functionality, still being affected by attempts to make it work, despite not even linking it in - thus, the troubleshooting part is actually not obvious to people not closely following the devel list and commit logs - and I can relate to that, despite being able to revert the offending commits.
Besides, the change was announced and introduced as being a "no-op", and the talks/commit logs later on could also make people believe that the problem is actually sorted - which does not seem to be the case.
Anyway, being able to exclude PUI from the compilation would indeed be useful - if only for benchmarking/regression testing purposes (think effect/glsl issues that are attributed to PUI incompatibilities) - just like it would be good to be able to disable Nasal scripting altogether for pretty much the same reasons: http://wiki.flightgear.org/Howto:Disable_Nasal_entirely
Thus, I really don't think that the motivation here is unreasonable
But like Thorsten said, adding/raising required libs is generally something to be handled carefully.