Submariner connectivity fails after a gateway failover with OVN-Kubernetes.

The failover creates stale gatewayroute and nongateway route CRs.
In the cluster where failover happens existing CRs have wrong next hops. In the other cluster the old CR is not getting deleted due to a missed endpoint delete event.

This leads to wrong routes being programmed.

To recover from this issue, the gatewayroute CRs can be deleted from both the clusters and then restart the gateway node.