In a large scale case of @titoiride, the checkpoint file has the wrong upper bound of the domain in checkpoint*/WarpXHeader lines 10 & 11.

The simulation uses:

• load balancing
• a moving window that starts and stops
• checkpoint-restart: after the 2nd restart it gets the bounds in the WarpX header wrong
• the issue appears with a restart after the moving window stopped

We still work on a reproducer.