Hi,
I'd suggest you looking into the logs of the elastic agent. It could be caused by a lot of reasons, like no agent auto-approval configured, network connectivity issues, etc.

I'm running into this as well, trying to figure out the cause. I have one pipeline working fine and one where the agent never seems to pick up the job and I can't figure out what's different. Same elastic profile. It starts the instance and the agent gets all the way to registering with the server and getting the cookie. After that, nothing and eventually the instance is torn down and another one started.

One difference I note is that I see [Should Assign Work] Processing should assign work request for job messages from the plugin in the working pipeline but not for the non-working pipeline.

Another difference -- the working pipeline was not in an environment and the non-working one is. Taking the non-working one out of an environment allowed it to work and the agent actually got the task scheduled to it. Still investigating, but this seems like a good clue.

After more testing it seems pretty clear the issue is that pipelines in a GoCD Environment do not get scheduled when using this plugin. Simply taking the pipeline out of the Environment will work around the problem. I think this must be a bug in this plugin -- the ECS elastic plugin does not exhibit this behavior, for example.

This plugin does not have GoCD environments feature implemented yet, it should not work.

The solution to it is rather simple: in "elastic profile" tab add the following string to user data field
echo agent.auto.register.environments={{ name of the desired env }} >> /var/lib/go-agent/config/autoregister.properties