VM to another VM FIP address (in different subnet) fails on second try

I have 2 VMs each in a different net. Each VM on a different compute.
In each repetition I do the following:
Create router and connect it to both subnets.
create 2 FIPs.
Associate each FIP to one VM.
SSH to one VM and ping to the other FIP from that VM.
SSH to second VM and ping the first FIP.
Delete the 2 FIPs.
Delete router (and its interfaces)

On the second cycle, the SSH to first VM failed.

In the attachments, you can find:
1. q-svc and Karaf logs
2. dump-flows,groups and ports from first VM.
3. dump flows, groups and ports from second vm.
4. Karaf models.
5. output log.
6. Test script.