>supervisor's research projects (lots of grunt coding)

Hobby projects that A) have no users so stability and usability is unimportant, B) have no one review code, or the results so correctness is unimportant, C) have no requirements to support it after the paper is written, so maintainability is unimportant all add up to shit general skills.

If a PhD has specific expertise you need then great, otherwise expect a mediocre entry level dev with a chip on their shoulder.