This sounds like given a coherent structure to “I want the prod code here to change the retry policy so pending retries are held in memory in this edge layer instead of the DB writers so load can be shed more cheaply” and it will usher things in from coding, testing, filing tickets, getting prod approvals, using the deploy tools to respond the containers, smoke testing the deploy, looking at Datadog or whatever to ensure the change is having the desired effect. All in one context, the tools and dev/qa/prod systems naturally being orchestrated within the same environment/ session. K8s is just the next to last 100 meters.