Currently, we have 256Mi as a memory limit for CO pod. I was able to hit OOM state and get CO restarted many times per hour (30-50 restarts). I reach this situation with 2 kafka clusters (3 and 1 nodes), KafkaConnect and KafkaMirrorMaker2 deployed in single namespace. You can reproduce it by setup same env a we do in MetricsST#https://github.com/strimzi/strimzi-kafka-operator/blob/master/systemtest/src/test/java/io/strimzi/systemtest/metrics/MetricsST.java#L294
We should extend the limits in examples or utiliaze memory usage in CO.