I've seen various posts claiming that the CABINET simulation in the KPA was the last stage in the signal chain, after the EFFECTS session, i.e. not as it looks on the front panel. But the following experiment supports that the signal chain is actually exactly as shown on the front panel.
Setup: Select any overdrive-type STACK with cab on. Set DELAY in EFFECTS to 'Legacy Delay' with 2 repeats,' Note Value 1' = 8/16, 'Note Value 2' = 16/16. Connect your headphones or a FRFR to MAIN out and listen to the STACK sound including CABINET. When you play a note on yer Guitar, you should hear the note plus 2 repeats: pling ... pling ... pling.
Preparation: Switch CABINET off and do the same. Now the three plings will sound differently, typically bright/fizzy. If not, select another CABINET, where the presence/absence of the CABINET simulation is very well audible. Switch CABINET on again.
Experiment: Put a finger on the CABINET button. Play *3* short notes on the Guitar now, quickly switching off CABINET after the 1st note and switching it back on after the 2nd note . The undelayed sound of the 3 notes should then be [normal] - [bright] - [normal]. Now listen to the delay: *both* repeats of the 3 notes will be [normal] - [bright] - [normal], like the original, even if the CABINET sim is back on while the repeats ring.
This more or less proves that CABINET sim is before EFFECTS, and that the CABINET sim output is fed into EFFECTS, as shown on the front panel.
Consequence:
a) there is no way to put simulated stomps into a simulated FX loop of the profiled AMP.
b) there is no way to put simulated stomps between the AMP and the CABINET. You don't do that in real world either.
c) "Monitor Cab Off" is rocket science, or uses an additional EFFECTS chain simulation. I do assume the former.
Additional, simpler evidence by Andrew d'Angelo in facebook: Put a fuzz into the X EFFECT and it will wipe out any differences between different CABINETS selected.