different live_update steps for the same image

In the #tilt channel, Robby Pocase asks:

Docker-compose use case - is there a way to have two of the same image running and have different live_reload options? I have two containers which include equivalent source code. One hosts a running application and one that just does the equivalent of jest --watch for running the application tests.

Due to a lib dependency, something is breaking --watch on sync, so I'd like to do restart_container() (or whatever the current recommended equivalent is) on the test container, but not the application container.

I've heard other versions of this. e.g., you have one big "mega-image" with all your services, and you invoke it with different commands in the k8s yaml

The best way to do this right now is to use dependent images: one base image, and then have images with different names that depend on it -- they don't add any layers, they just add the live-update steps.

I don't have a good idea for this now, but wanted to log it in case there are other use-cases I'm not thinking of

Start by reviewing Tilt's live_update behavior and how dependent images are currently represented. Compare the Docker Compose case with the mega-image and Kubernetes YAML examples described in the issue. Done means there is a decided, documented way to assign different live-update steps to equivalent images without unnecessary image layers.