The introduction of goroutines seems to have caused some race conditions in the VPA in #7992

This can be reproduced by running the current master VPA in a cluster, and creating 1000 VPAs in quick succession.
The VPA recommender pod crashes

(sorry for the lack of details, I'm busy working on this)
/assign
/area vertical-pod-autoscaler