Rudder

On large installation (10k+), the memory usage when writing expected reports is absurdly high; same when fetching data
We should batch these queries, because it means we are taking everything, serializing them all, and putting all theses serialization in big ldap/sql query, so we are effectively tripling (+ side effect) the memory usage at these operation

Hypothesis: batching by 200 to 500 would be sweet spot because it keeps the memory at bay, while still transfering enough data
To test this hypothesis, we need to make a parameter that can be changed (either by api, or by configuration parameter) and test several scenarios

Note: version prior 4.2 where not impacted, because we didn't manage data the same way