Previously, creating or updating entities would trigger lots of reindex calls back to entity-api (also depending on the provenance graph). And several times such operations caused the entire VM to hang or become extremely unresponsive. Neither SSH access nor serial console worked. A forceful VM restart became the only solution.
To ensure no single container can monopolize the VM and cause hangs, we should set the CPU and memory limits.