LTM HA meltdowns with nonsensical self IP

This turned out to possibly be bug 658716 as reported in 13.1.X release notes (https://support.f5.com/kb/en-us/products/big-ip_ltm/releasenotes/related/relnote-supplement-bigip-13-1-0-4.html)

Support said this bug was actually introduced in 13.0, so we had to downgrade to 12.X, which was a huge PITA because the config files are not downstream-portable. Be careful with 13.X

Jun 27 14:05:27 ext-lb02 warning mcpd[27298]: 01071aea:4: CMI heartbeat timer expired, status: 10.10.16.1 Followed by HA disconnecting and daemons restarting, etc.

This happened on only one of two pairs of HA vCMP guests in a cluster. We left the other pair of guests and the hypervisor alone for now but downgrading fixed it.