I’ve done this for host integrations a few time - by from the host, call Python code that “opens it up” for a remote debugger, and then connect to it.
I use Pycharm and for example the logic here is very close to what I’ve used to run the debugger in Maya, Houdini, etc. whilst running in dev mode (so it runs of the code I also have open in my IDE) This works surprisingly well for any of the client side work (aka the “pipeline layer” in Mustafa’s post).
For the docker, server and backend I’m not sure - that’d be up for @martin.wacker to maybe provide some pointers on and I’m clueless.