For performance and usability reasons, we should consider limiting the number of event sources we use for autogeneration. This could also be configurable for the user to give them some flexibility. Furthermore, we could consider having the sample sizing for trace and sample value lookups configurable too, for the sake of tweaking performance.

Some further context: https://github.com/camunda/camunda-optimize/pull/2276#discussion_r448251591

There is theoretically a chance that the current unlimited source count could result in reads exceeding the set limits, so having some safeguard against that would also be beneficial.