Movement compensation of CTF data by Maxwell filtering *without* external interference suppression?

I would like to do movement compensation of CTF raw data (several runs of length 15-20 Minutes), in order to concatenate epochs within them in sensor level for decoding analyses. I still want to be able to apply the CTF gradient compensation, since it appears to me to give superior noise suppression over pure tSSS in my dataset.
I already extracted the continuous head position with mne.chpi._calculate_head_pos_ctf (#4088) by @bloyl.

However, from the old discussion in #2277 (@staulu, @larsoner, @wronk), I understand that

1. it is not possible to perform CTF compensation after SSP has been applied (by the way, also, when only applying it to the data channels?), and
2. it is not yet (?) clear, whether SSP can be applied to the gradient compensated data.

So, my questions is, if setting st_duration=None in maxwell_filter will prevent maxwell_filter from doing external interference suppression and only do the movement compensation? And whether that procedure might actually be OK (for leadfield computation etc.)?
I tried to understand the code (thereby encountering some other questions/confusions), but unfortunately cannot answer this question with confidence myself.

A subsequent questions would be, if it is better to first apply movement compensation and then the gradient compensation or the other way round. (My intuition would opt for the first way.)

Thanks in advance for any thoughts/help/advice!!

Start with mne/preprocessing/maxwell.py and the existing _calculate_head_pos_ctf path, then review issues #4088 and #2277 for prior decisions about compensation and SSP. Establish whether st_duration=None supports the requested movement-compensation workflow and document the intended ordering and completion criteria.