It should be generally safe. TO be extra safe, you might want to shut down the workers to let events queue up. Then do the upgrade, then start up workers on new code.
It’s hard to make them 100% graceful with tagged releases though, so this might not be perfect every time and you’ll get some intermittent errors potentially during ingestion.
Also to minimize the effects, it’d be a bit better do 8.10 -> 8.11 -> 8.12, etc. I know that’s much more effect, but less likely things will error in between.
Ingestion from events being sent from the web processes. Before being queued and processed. Generally this should be fine during an upgrade without downtime.