Vmstat/Top reporting inaccurate CPU usage, potentially related to microcode version

We have seven similarly configured and loaded ec2 instances running centos7. Over time, three of these instances have started reporting cpu utilization around 0.1% in vmstat or top, while CW metrics shows 40-50%. In the beginning, and for the remaining four instances, vmstat/top roughly agree with CW metrics. While looking into this, I realized the three that are reporting bad info have newer microcode versions. I could see that the newer microcode version could potentially be a sign of some other changes (newer underlying hardware, etc), so not saying it is due to microcode.

This is throwing off our monitoring software. How can this be fixed?

from /proc/cpuinfo on instance showing wrong cpu usage: microcode : 0x5003501 from /proc/cpuinfo on instance showing expected cpu usage: microcode : 0x42e