For completeness sake. I will drop some findings here but I will start with this.
TL;DR: Returned to original batteries → Problems seem to be mostly solved.
Batteries:
I saw what seemed brown outs, and I imagined that the batteries (7565121 batteries as discussed in Getting more than 10 hours of battery runtime from the RK3588 Pocket Reform - #13 by grimmware ) were the only change from the “factory” setting. I recovered the default batteries and both the brown out and the keyboard issues seems to have been solved.
Some Traces (before battery change):
I was commenting with @minute in IRC about the problem, and to double check the cable I crimped a new one to test if there was any change at all.
With the new cable the problem persisted, so either I crimped it terribly (which could be) or the problem lies somewhere else.
I wanted to share here the USB tracing that I captured and saved in a format valid for pulseview/sigrok. (https://app-share.mailbox.org/appsuite/api/share/04f05e2f012695414b67deb126954532a4abbda9afd6c11c/1/8/NDQ/NDQvNQ).
On a small analysis I can see the bus halting for some time periodically, and afterwards there is more activity, I guess is the re-negotiation of the firmware with the system. This happens soon enough after powering on, less than 2m in.
Summary:
I will keep an eye to dmesg for more problematic traces. I can still see sometimes the dreaded reset full-speed USB device number X using echi-plaform but it does not seem so prevalent.