Currently, we are testing just actual latest version of Kafka, which we are supporting, during our regression pipelines.
It's because of the time (and resource) that it would consume when we would run regression pipeline against every Kafka version supported inside Strimzi.

It would be great to have some kind of smoke test, which would, during the regression pipelines, just deployed Kafka cluster, did some basic operation (such as sending and receiving messages) and then just delete it. This test would run for every supported Kafka version.