Oran Agra

we're aware of that (command stats are updated when the command actually runs / executes it's logic). the notes at the top include > One of the changes in this...

let's merge this one as a bug-fix, and the other one as a new feature..

I certainly agree with the [optimization](https://github.com/redis/redis/issues/2702#issuecomment-1475351513) (`SCAN {user}-*`), is that what you meant in your second bullet in the top comment? we can't have it scan the entire slot in...

@hpatro any news here on my last request?

ok, so that config takes only some 20ms for a 80k clients. sounds ok to me

Full CI: https://github.com/redis/redis/actions/runs/3618307056

using `src/redis-benchmark -t set -r 100000 -n 10000000 -P 10` i can see 5% throughput improvement with default configuration (`save` and `maxmemory-clients` disabled)

@hpatro @madolson i implemented my notes since i would like to include this in the next 7.0 release. please have a look and let me know if i can proceed.

@filipecosta90 ~~you should probably compare the calls for updateClientMemUsage with updateClientMemoryUsage, not updateClientMemUsageAndBucket. i.e. all the places that called updateClientMemUsage now call updateClientMemoryUsage, and the AndBucket is almost never used.~~...

conceptually approved in a core-team meeting.