I almost always have this issue under load (and therefore heat), but it can be a bit obstinate about fixing itself.
I have also had issues with the speaker connector/audio board interface since replacing the bottom cover, which I resolved by annoyingly precisely tweaking the spacing of said connector. Post here.
Notably, my new bottom cover is not cracked, but I can trigger intermittent connectivity with the speaker pogo pins (or rather, could, before inserting just the right thickness of cardstock into the slot) by applying light pressure (literally resting my hand) to input cover in the right position.
After spending far too much time tinkering with both the keyboard and speakers, my guess is that the flexibility of the bottom plate, combined with the sometimes squidgy placement of the screws (I noted that the audio board levers a bit oddly due to the placement of its one screw relative to the pins that connect it to both the speaker connector and the main board) gives more variation in the pressure of the pin connections than they are toleranced for, leading to intermittent connection failures with temperature and mechanical stress.
Since the connection is also bad if there is slightly too much pressure, I am not sure if there’s an extremely straightforward fix.
I have a support ticket open that I really should get back to, but I’ve been out of country and then ill, so it’ll happen when it happens.
Edit:
tl;dr, I think this and the speaker problem are the result of the pogo pins not having tolerances adequate to the combined flexibility of the bottom cover and PCBs (which are also bendy), especially when the whole system is subject to thermal cycling.
To be fair, it’s entirely understandable that that kind of emergent issue wouldn’t show up in design or prototyping.