(Confirming @guillaume-blaquiere's comment) I just had the case where my containers (min instance = 1) lost their connection to an Atlas mongodb cluster and would not die but rather return 500 to every request.

Using the console, I performed following steps:

1. click on my Cloud Run service
2. click "(pencil icon) Edit & deploy new revision"
3. leave all settings as is, including the container URL
4. click the Deploy button

This created a new revision quite quickly and directed all traffic to it. My instances started from scratch and connected properly to my mongodb cluster that got back online in between.