"draining" historical nodes

Hi Druid users,
The clusters I’m running periodically need resizing, both upward and downward, and while I find adding historicals to be straightforward, scaling down the historical tier(s) is not as simple.

I’m curious if others are facing this challenge? Is there any guidance for “draining” historicals, similar to how middlemanagers can be disabled for zero-downtime replacement?

–Ryan