Miguel Ángel Ortuño

In the end, I decided to adapt chunks pool to contain slices with a total capacity of `10k` chunks per slice, rather than managing plain `storepb.Chunk` objects. And now benchmark...

> Or is it just a big but not excessively huge approach? That's the idea. Though still an arbitrary number, susceptible to be changed. I'll add some comment to clarify....

Closing. [XEP-0313](https://xmpp.org/extensions/xep-0313.html) support already available starting from v0.62.0

> Do you think it's worth and feasible putting this behind a feature flag to make rollback easier in case any unexpected issues appear? I believe it makes complete sense...

Added new `-query-frontend.use-shard-active-series-zero-allocation-decoder` experimental option to enable/disable this feature at will. _@flxbk @Logiraptor this is ready for review now_

_pinging @flxbk_ ready for review again 👍

> I do worry about debugging this if it ever breaks, so I added a fuzz test which compares this PR to the stdlib json result. The first failure is...

_pinging @flxbk @Logiraptor for a new review_

As a reminder, don't forget to add admin client cache specific [parameters](https://github.com/grafana/backend-enterprise/blob/b6875a9e81a7a21d7fafa67c9e5f522f3c19e234/docs/enterprise-metrics/docs/config/reference.md#admin_client) to the `admin_client` section in the `values.yaml` file to make it point to this new component. These are...