PFFB is not a panacea. The FR examples in post #93 show that the variation of 10 kHz output with 4 vs. 8 R is minimal even without PFFB, so what are you hoping to gain from PFFB in your application?
Bruno Putzeys' (of Hypex UcD and ncore and Purifi fame) whitepaper about the F-word explains the issues at stake quite well. PFFB will lower output impedance and hence load sensitivity. Feedback will act as a noise shaper, so it will lower low order harmonics at the expense of higher orders which are usually more unpleasant. If PFFB is simply used to obtain passable distortion performance from questionable passive components such as the output coils, it is no good. If you can live with the load sensitivity, i.e. if your tweeter is crossed over below 3 kHz and has no impedance peaks above, I simply do not see why you should go for a half-baked PFFB.
TPA chip based PFFB should really be done only with excellent passive components and sufficient loop gain and bandwidth. The loop gain around the TPA chip is limited, so either excellent output filter performance can be improved upon or a poor output filter can be camouflaged to measure better on some popular metrics but really you just end up shifting the junk to higher harmonic orders where it might be more audible.
I am not sure if the champions (Topping, 3e) use only a PFFB wrapped around the TPA chip and some clever gain structure upstream or more likely have some kind of composite approach where a separate and fully analog op amp is part of the feedback loop around the TPA chip and the filter.
As for the HP filter (and I quite like the idea of not asking your mains/satellites to play LF), if that is what you are after, there are many ready-to-run line level modules availeable on Aliexpress, or you could roll your own with the help of ESP projects in Oz.