Since updating to the latest RM (3.4.32) I've noticed the virtual knobs behave slightly differently. In general they now seem more fiddly/less predictable, but the main problem is that on the parameters that start from 0 and go to 10 (e.g. bias/sag/comp etc) they now move straight from 0 to 0.2 - even using shift+mouse you can't get to 0.1. May seem like a small gripe, but often 0.1 is all that's needed! Similarly, at the top of the range, they now go from 9.8 to 10 - less of a problem for me, as usually I prefer very small amounts of these parameters, but still...
And I realise you can still access these values using Kemper's physical knobs - but I've always found these fiddly to work with due to their non-linear nature - I can see the thinking behind them being non-linear, presumably so you can quickly move by large increments using fast turns, fine tune with slower turns - but in practice, it's just fiddly... would much prefer linear behaviour, like a real analog dial. But, that's a side note - so long as the virtual knobs in RM behave predictably and allow you to access the full range of values for each parameter, I much prefer to use these anyway.
And one other very minor gripe, but something that has made RM just slightly less convenient to use - previously, the mouse wheel moved the values in equal increments of roughly 0.3 in any direction - when testing a profile out, I'd typically set rough values using the mouse wheel just to get a quick idea if adjusting said value was going to work. Now, on the positive/negative dials, if you turn it in the negative direction it behaves much as before, but in the positive direction (using mouse wheel), it goes from 0 to 0.1, to 0.4... I warned you this one was a minor grip 
But still, was much more useful before to have it go 0 / 0.3 / 0.5 / 0.8 etc, and to have it move in the same sized increments whether you turn it negative or positive.
Summary - could the "virtual knob" behaviour, go back to how it was before, or at least be tweaked to follow some other logical/consistent/predictable pattern? But mainly, we need to be able to access all the possible values using the RM knobs - I think I'm not alone in far preferring RM to tweak profiles with than the Kemper's physical knobs.
