The Intel SAS controller sometimes has some issues with Linux, unlike the Intel SATA ports which are rock solid, if you can try for example using an LSI HBA instead.
Looks more network related,try to simplify you lan config as much as possible, also that doesn't show the beginning of the trace, but if using custom IP addresses for docker see here:
Yes, I wanted to see a syslog with multiple parity checks, before rebooting, that Marvell controller doesn't usually cause sync errors, but it's a possibility, and like mentioned not recommended anyway.
Any controller with SATA port multipliers is not recommended, Marvell controllers are also not recommended, even without port multipliers, if 5 ports are enough you can get a JMB585 based controller, if you need more look for an LSI HBA.
The extra controller you have is a 2 port Asmedia controller with SATA port multipliers to give the 6 or 10 total ports, these are a known problem and will cause constant timeouts, you should replace it.
It does, but if I do the procedure like you specified above I still get the same result, i.e., "wrong" disk after a reboot and an unmountable damaged partition.