Evaluate impact of db connection pool on throughput and performance

[FloThinksPi]

Context

Currently each CC consumes 20 db connections by default. For Amazon RDS(postgres) the connection limit is 5000 LEAST({DBInstanceClassMemory/9531392}, 5000) see https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/CHAP_Limits.html

When also taking CC workers into account one can estimate rughly 200 CC instances can be started and no further scaling is possible. Other hyperscales have similar restrictions.

Can we tweak this value down safely to increase the number of CC instances and thus allow for higher throughput and load without optimizing the implementation in particular.

Acceptance Criteria

• [ ] We have a document showing how changing this parameter influeces 1. troughput and 2. performance