If a VM migration takes a long, the descheduler may attempt to evict more VMs in parallel. We've seen up to 5 evictionRequests at one point of time:

I0617 06:58:50.396955       1 descheduler.go:403] "Number of evictions/requests" totalEvicted=0 evictionRequests=5 

Since KubeVirt allows only 2 concurrent migrations per node by default, additional migration jobs remain stuck in a Pending state, waiting for available migration slots

When using the DevKubeVirtRelieveAndMigrate profile, we should limit evictions by default (set 'evictionLimits.total=2') to match the KubeVirt migration limit and avoid unnecessary evictions.