Grain balancing across silos

[chocvanstraw]

Several years ago, I built an Orleans based system that had a fairly constant number of grains (around 3000) with reminders distributed across 3 silos running on 3 VMs. I noticed that when one VM was brought down for maintenance all the grains would move to the other two VMs leaving the 3rd one empty when it came back up since very few new grains would be created.

I'm looking at Orleans for a new project and wanted to know if there was any solution to this? Since my plan would be to host it in Azure, I would want to make sure the VMs or App Services (if I go that route) are being utilized.

This is similar to the following discussion #3336 from back in 2017 and I was curious if anything has changed since then or if the original guidance still stands?