While working on T357380: Degraded RAID on mw2442 I realized a strange looking RAID controller config (for each disk there is one "RAID-0" virtual drive created which then is used to create an mdadm software RAID) which turns out to be the same on all 32 hosts of that batch.
As of the procurement task T325215 those systems (Config C-1G) should not have RAID controllers at all, so I assume something went wrong during procurement as well as during provisioning as T326362 does not request HW-Raid config.
This is not ideal as it makes those 32 hosts different from the others and it does require extra care/extra steps in case of disk replacements (see T357380#9575876). We should probably re-provision those hosts with the RAID controllers configured in HBA mode or have the RAID controllers removed (if that's even possible and it makes sense to keep spares).
Opening this task to figure out what to do or at least keep the information around that probably the following steps are needed when a disk is replaced to make it appear in the OS:
megacli -GetPreservedCacheList -a0 megacli -DiscardPreservedCache -L'disk_number' -a0 megacli -CfgEachDskRaid0 WB RA Direct CachedBadBBU -a0