Read the previous 2 articles for context.
The issue was (apparently) different SSD models in the same RAID which the controller didn't like. The RAID has been recreated with 8 identical SSDs, this time in RAID5, which allows for better redundancy than RAID0 (some redundancy > no redundancy). I did the same test again, filling the disk up with random data, and it completed successfully.
This means that the problem should not happen again, at least not any time soon.
🫡