Steps to Reproduce

1. Have a service up and running on CRC
2. Update the deployed service to a newer version that fails on readiness probes
3. Update the deployed service again to a version that fails on readiness probes

Actual Behavior

As the first deployment (2) fails on readiness probes, the original service (1) doesn't get deleted and gets requests if exposed. When the second failing deployment (3) happens, the working original service (1) gets evicted and we end up with no functioning pods and service downtime.

Expected Behavior

The original working deployment (1) should outlive the first failing one (2) after applying the second one on it (3) so there is no downtime even if failing services get deployed on top of each other.

Business Impact / Additional info