@HADES Thanks for giving such a clear explanation.
The driver only exposes the hardware PPEC implementation. And this one kills guide pulses as it overloads axis speed at constant interval using speed values stored in microcontroller eeprom (after training).
As long as you have continuous axis speed setting in a driver, PEC is better (and in a simple manner) performed using a software implementation. That may be done at the INDI telescope class level, but I wonder if guiding software should not be aware of this PEC correction. It seems it is best if they do the whole thing.