Production memory reached critical capacity after some application processes failed to connect to a core database, and memory was not released cleanly. We have made alterations to the frequency of open connections, designed an initial task to reset connections regularly, and increased our alerting levels to ensure no re-occurrence of this incident.