For exploratory, my take is rotating every release is usually OK. You need the area expertise for good exploratory but don't want them to get jaded. Since they roll their own experiments it doesn't happen as fast.
But scripted is far different. Using manual testers for scripted regression is a bad idea for all kinds of reasons, but you honestly should never use them more than once in a row in the same area. We're just too hardwired to autopilot or cut corners when doing scripts once they're learned. But once is OK, so it's novel, and when they forget later it can be OK again until they learn the script again. It's why crowdsourcing has actually worked pretty well in that area.
But really, just don't use scripted manual. It doesn't work. Put people on exploratory, put scripts in automation, don't pretend either can do it alone.
I don't think it's a function of career length, though, and that threatens to be a bit ageist to be frank. I'm quite sure I found more and better bugs the length of my primary-QA career, so long as I cared about the job and didn't have to recite a script. Burnout times were different, but those aren't age-bound.