You can record in the Performance (audio) recorder to create .wav files that are saved on a USB stick. You cannot, however, use this mechanism to "bounce" because you cannot play a .wav file while recording. Therefore, you would need to take this .wav file and assign it to a key (using a drum key would be easiest). Then you can trigger the .wav file by pressing and releasing a key. If the .wav file you recorded has been setup properly as a drum key with proper settings then the sample will continue to play to the end of the sample without your having to hold a piano key down.

Once you've placed the first pass into a drum key - you can go back to the performance (audio) recorder and do a second pass. "Rinse and repeat"

The second pass you can replace the drum key you created in the first since the new recording has the original audio plus added parts - so you can replace the old drum key .wav file with the new one.

This is how you could do this using just the keyboard itself and no external gear (except a USB stick).

The more elegant way to do this would involve a computer of some sort (including tablet as a form of computer). Using a DAW to record audio would more easily handle this with less steps for each pass. I understand this isn't what you asked for - but it's worth mentioning the merits (and downsides) of the alternatives while still respecting there are likely good reasons for your preference not to use this path. If you don't keep the keyboard in one place - lugging around a tablet (or desktop/laptop) and adapter(s) (USB, power, etc) may become cumbersome.

”Bouncing (or exporting) is how your DAW turns your project into files on your hard drive. The term bouncing comes from the analog era. The track count is a hard limit on tape machines. ... But today, bouncing usually means writing the final mix of your song to a stereo audio file.”

You are asking about MODX/MONTAGE, but you posted in the MONTAGE forum:
You have 32 audio bus outputs on the MONTAGE; 10 audio bus outputs on the MODX.

If your workflow includes recording MIDI Tracks first, then rendering audio, you have several options and methods available. You don’t have to wait to create the final mix, you render audio whenever you desire… as you mention: to save polyphony.
You can begin MIDI Tracking in the synth or in Cubase, and once you have corrected any mistakes, you can render that data as an Audio Track in Cubase.

You can playback the Audio Tracks while you continue to use your MODX/MONTAGE to create additional MIDI Tracks.
The nice thing about this workflow, is you keep the original MIDI data ‘muted’ — and playback it’s audio version. This frees the synth hardware for a completely new recording. If at anytime you decide to redo a track you can trash the audio, recall the MIDI data, fix it, and render a new Audio Track.