> Once you actually sit down and come up with a project plan with your PMO [...] The risk of regressions, the refactoring, the retraining, the politics, etc are hardly ever worth it.

You are entirely correct, of course ... except that much of the management class simply does not care about any of those things.

Not to mention the sizable contingent of engineers will repeatedly get suckered by the pitch of "Just migrate all of your stuff to [shiny new thing] and all of your [reliable old thing] problems will go away" (a.k.a., "engineers with management potential")