I will look in to this and update when I find out if it's an option. In the meantime, our lab Compact running the same kernel and same python application failed today with the same symptoms. LEDs look normal, but device is unpingable. I connected a monitor, and it is scrolling stack trace info. Unfortunately the videos I've taken aren't able to capture the text as it scrolls, so I only have an image of the end of the message.
A few lines I could make out from the video:
rcu: INFO: rcu_preempt detected stalls on CPUs/tasks
rcu: rcu_preempt kthread starved for 6305 jiffies
Tasks blocked on level-0 rcu node (CPUs 0-3): P13/2
task: irq/66-dwc2_hso state: S stack: 0 pid: 87 ppid: 2
rcu: RCU grace-period kthread stack dump
task: rcu preempt state: I stack: 0 pid: 13 ppid: 2
