I'm not very familiar with those controllers but after taking a second look looks more like disk2 is causing the HBA related timeouts, what disk(s) did you last replace before you got read errors from disk2? Do you still have those old disks intact? Was anything written to the array after those disks were removed?
Downgrading to v6.9.2 will also work, but you can download the v6.11.1 zip and extract all the bz* files to the flash drive replacing existing ones, then reboot.
Pulling the disk won't fix anything since the emulated disk will have the same problem:
Nov 6 17:25:09 1mehien kernel: BTRFS info (device md13: state EA): forced readonly
P.S. disk14 is also showing issues now:
Nov 6 17:13:46 1mehien kernel: BTRFS info (device md14: state EA): forced readonly
For both now.
Doesn't look like an Unraid problem:
Nov 6 20:31:56 ser kernel: nvme nvme0: Device not ready; aborting reset, CSTS=0x1
Nov 6 20:31:56 ser kernel: nvme nvme0: Removing after probe failure status: -19
IIRC another user had a similar intermittent issue with the same or similar Samsung device, try power cycling the server and/or downgrading back to v6.11.0 to see it's still detected.
Do you mean you are suing the disk share for transfer 1? Disk shares are always faster but a user share should be much faster than that, diagnostics might show something.
We usually recommend rebuilding, as long the emulated disk is mounting and contents look correct, other option is doing a new config but you'll need to do a parity check, and that will take as long as the rebuild.
Do you mean the pool does not mount as was the problem for the OP or the server does not boot? I guess the latter since the unassigned NVMe device is not using the same filesystem.
Constant ATA errors form the cache device, they look more like a connection problem but SMART does not look good, run an extended SMART test and/or replace/swap cables to rule them out.
User shares always have some extra overhead, creating subfolders will help, another thing that should help a little or a lot, depending on the hardware used, is to disable the security mitigations.