The reason it requires reduction to 11MP and to use the video pipeline instead of the Expeed pipeline is due to sensor readout time. I don't /fully/ understand it, but the sensor readout data on the Rolling Shutter Project shows a Nikon Z8 would max out at ~22fps in full-frame RAW, so no doubt they reduced this to 20fps to give them a margin of error. Reducing to the DX frame size reduces the amount of data required to be read from the sensor, which changes the time it takes to do a readout. The Z8/Z9 are 45MP cameras, so just doing a naive bit of maths, you could expect ~80fps if you applied the same pipeline at the DX frame size, but given that there is already a video pipeline and they needed it to be capable of 120FPS, they already had their answer.

both video and photo pipeline is expeed. it does not go into DX mode for 11MP, its full sensor size. its unclear(atleast to me) if it does some form of line skipping or if its oversampled fully.

but the video mode supports full 8k60 atleast, so only a very tiny crop.