I’ve spent more time fiddling with the LTE-M XBee and I’m seeing the same results.
It’s a hardware problem, not software; the modem really does stop responding. I wonder if it’s that there’s not enough capacitance along the power line causing very short voltage drops even when it’s tied right to the battery. The user guide for the LTE-M XBee specifically warns that the main control board on the XBee can become unresponsive after power dips and will need a full power-cycle to get it back. Oddly, though, I’m having it start responding again without actually power cycling it.
I have a Skywire development board that has more capacitance on the power line. I’ll fiddle with it and see if I’m having the same issues.