Thanks for testing and reporting back.
That confirms the regular activity was being caused by zimaos-welcome@tty1.service, not by an incompatible NVMe.
After stopping the service:
- the fixed 2–3 second activity disappeared
- idle CPU dropped from around 4–5% to 1–2%
- the remaining writes are small and irregular
- the remaining activity is consistent with normal journaling and running apps
The service repeatedly checks the network, hostname, uptime and gateway configuration to refresh the directly connected console display. On your system, that continuous polling was enough to create noticeable NVMe activity and CPU usage.
The service will likely start again after a reboot, so I suggest leaving it stopped for now and reporting this behaviour to IceWhale as a ZimaOS 1.6.2 issue. I would not permanently modify or remove the system service until IceWhale confirms the intended fix.
This would also be a useful case to test with ZimaBrain CE. It can capture the system-drive layout, NVMe health and temperature, active services, failed units and other local evidence in one report. If you install it, please let us know whether it identifies anything relevant or misses anything in this case. That feedback would be very useful for improving it.
Please also let us know whether the NVMe and CPU temperatures reduce after the system has been idle for a while.