I remember reading a thread recently where the use of PLA to support PETG and vice-versa was discussed. Several users found that parts ended up weak at the layer where the filament change happened and there was discussion about contamination being left in the nozzle and affecting the subsequent printing. Interestingly, the observation was that using PETG to support PLA works but using PLA to support PETG doesn’t fare well 'cause the PETG ends up weak at the filament change layer. IIUC PETG needs the previous layer to remain warm in order to fuse well, so now I wonder whether the issue there wasn’t contamination but the same you observe here, which is too much cooling of the previous PETG layer…
Something you could test is to construct a print using all PAHT-CF with an auxiliary structure (similar to the prime tower but perhaps larger) that has a time-consuming layer somewhere to simulate the filament change without actually changing filament. E.g. minimal infill for almost all the tower but one layer solid and tweaked to print at the lowest possible speed so it takes about the same amount of time as your filament change. If that produces the same weakness you have your proof… (I can also understand that you may not need any further proof at this time…)
Another idea: would it be possible to perform an ironing pass over the layer previous to the filament change to reheat it? Dunno whether ironing can accomplish that, I’m just wondering…
Yet another idea: observe the print with a thermal camera to verify the hypothesis…