how to handle discovery downtime due to agent taking time to sync endpoint slices

[leasonliu]

We are using Submariner 0.17.3, which has a known issue: when the lighthouse agent pod restarts, it takes significant time to sync all endpoint slices from the Submariner broker. With our large number of endpoint slices, the sync takes at least 8 minutes. We know this is fixed in 0.20, but we're not ready to upgrade yet. We want to apply some fixes/improvements to our forked Submariner and deploy it. What can we do to avoid downtime during Submariner releases?