Comment by valyala

Yes, there is a solution for masking the read latency at object storage - to run many readers in parallel. I tweeted about it some time ago - https://x.com/valyala/status/1965093140525715714