Worker node goes into a not ready state in OpenShift 4

Issue

• Load average is very high, over 300 and these are ~50 cpus/threads workers.
• Tons of "x process is being blocked for 600 seconds" messages reported to dmesg, so seems there is a problem from a kernel perspective to pair a process with a cpu to run tasks.
• Many process in "D" uninterruptible state, so waiting for something in a hung state.