Monitoring Fortinet cluster - strange and inaccurate utilization numbers

Hi all -

For the last couple of weeks, I've been getting very strange utilization numbers on our Fortinet boxes. We run a HA cluster with two Fortigate 600Cs.

We've made no changes of which I'm aware on the NPM or switching side, and have certainly not changed the Fortinets - but I regularly see numbers like the following.

These numbers are not accurate - by a long shot. It almost looks like I'm looking at an inaccurate 32-bit counter that flips too fast to get an accurate number, but all interfaces are set for 64-bit counters.

Port 7 is the HA heartbeat port and has almost nothing going across.

Anyone seen this before when monitoring Fortigate clusters?

Thanks in advance for any info.