Stop node whilst array is rebuilt or risk disqualification?

[Alexey]

You may try to use this approach:

Windows disk check found.000

You may try this to recover pieces: in conjunction with you would get a script for PowerShell (just use ps1 extension instead of sh) to recover pieces. I guess you would need to change permissions for these files to match permissions of other files in the blobs folder.

However, seems either this controller is misbehave, or you do have disks with a higher bitrot rate. Since it’s a hardware controller, it doesn’t able to recover corrupted data and you see a result.
The best case for this server seems to use disks separately, in this case you would lose only part of the common data, not the whole.
The other options are:

• use Linux or freeBSD (TrueNAS maybe?) and zfs, the adapter need to be switched to handle disks separately, not as RAID, or remove it and connect to the motherboard directly (I do not know, what’s better);
• upgrade to Windows 2016/2019/2022 if possible and use ReFS, but again, disks should be exposed to OS separately. This option is less robust, but maybe ReFS is ready for production this time.