Thank you for taking the time to look into the CI pipeline and for the proposed improvement! 🙏

Closing this one as the CI workflow has since been reworked directly on master and now covers the same ground:

• the "copy config" step already does a docker restart + a /ping polling loop (instead of sleep);
• a ClickHouse Keeper config was added so ON CLUSTER tests can run on the single-node image.

One important note for the future: the ClickHouse server images do not ship curl — only wget. The health-check/poll here uses curl, which was actually the root cause of the CI being red since April (the service container never became healthy). On master the health-check now uses wget --spider.

Really appreciate the contribution and the investigation — thank you! 🦫