I don't have the manual handy, it is in storage, but that bias sounds a lot smaller than I would expect. 86kHz or 100kHz shouldn;t matter. Yes, one might be more optimal, but either would make it record or erase. Your problem is gross, not a nuanced frequency variation.
4/8 channel difference shouldn't matter. I'd be surprised if the bias oscillator and drive are different between the two.
If it comes and goes, I would suspect an R/P switch contact or a solder joint somewhere.