Problem is disk2, it's failing:
=== START OF READ SMART DATA SECTION ===
SMART Health Status: FAILURE PREDICTION THRESHOLD EXCEEDED: ascq=0xfd [asc=5d, ascq=fd]
Do you have another spare?
Up to date in the sense that now there's a SMART report for disk16, syslog is always lost after a reboot, disk16 SMART looks OK, though it looks like a white label drive, don't really recommend those, run an extended SMART test on disk5 and post new diags.
Syslog cuts off August 23 so we can't see the errors, but you should run an extended SMART test on disk5 since it's logged there as a disk problem, there's no SMART for disk16, power cycle the server and post new diags after array start.
P.S. a single stream iperf test is much better to give an idea of the actual transfer speed, when you copy using one windows explorer session it's a single transfer.
Currently only the WD SSD is in the pool, stop the array, unassign the Samsung NVMe, start array, stop array, re-assign the Samsung NVMe, start array and that should fix it, in doubt post new diags.