A RHCOS based on RHEL 9.2 (OCP 4.14) which had the nvme_tcp storage driver loaded, stopped to load it after an upgrade to a RHCOS based on RHEL 9.4 (OCP 4.16).

It did not happen on all RHCOS nodes, just a few of them.

Customer reported this on different OCP clusters, not a just one.

Upgrade took place from 4.14.25 to 4.16.34, customer didn't notice the problem while temporarily being in 4.15 (probably because it is still based on RHEL 9.2)

I've tried to reproduce this, in my case, a fresh OCP 4.14.25 loads the nvme_tcp storage driver, and after upgrading to 4.16.38, the drivers are still there being loaded.

Nodes in customer environment needs this driver because their NetApp CSI needs it to provision persistent volumes.