Repeat experiments with misconfigured SUT
Occasionally experiments fail in terms of producing wrong results, for example, because not enough partitions or Kafka Streams (or any other SPE) instances are created. We want to discover such situations at runtime (instead of looking at our Grafana dashboards) and restart those failed experiments.
This depends on, or is at least related to, #12.
It will be way easier to be implemented with #46 (closed), and might be even required for #48 (closed) and #49 (closed) for correct results.