Changed the code around a bit to make sure I gather all the logs generated by a single rip in the most simple way (using journald).
This allowed me to also support a "debugcam" - recording everyting happening on the robot workspace when a rip is executing. This uses a webcam working with ffmpeg to write a low-res low framerate video into the storage directory.
During this I discovered that ffmpeg can't record to VFAT because of a weird EINVAL returned by the kernel when trying to openat(). Discovered this using strace(), had no inclination to debug this further so I just changed the storage filesystem to be ext4. Reduces compatibility with Windows but meh.