We just ran into an issue, where reads were blocked due to

healthy-read-lag

limit being in pending batch writes. In our situation, we have a lot of webshops where we'd much rather give them content, that is a little outdated rather than breaking the search altogether, by not returning anything. We are considering increasing

healthy-read-lag

substantially from the default (1000) to avoid this in the future. Anyone running with a very high

health-read-lag

who experience issues with this?

It's fine if you are okay with stale updates. Too high a

healthy-read-lag

can sometimes mask underlying issues. As long as you do have some monitoring around it it should be fine.

OK, so it's not like a high amount of pending writes depletes memory?

No, these writes are bufferred on disk. There will be some memory usage but it's not that all of those are held in memory fully.