Very slow memory leak in API containers

[WadeBarnes]

There appears to be a very slow memory leak in the API containers. Here is a 2 week memory profile from the production OrgBook environment in the OpenShift cluster showing the pattern for a number of API pods:

The sharp drops are were container instances in the API pods have gone over their memory allocation for too long and have been killed and restarted. Each API instance is limited to 768Mi of RAM and multiple instances (6 in prod) are run and the number can scale to a maximum of 16.