Reporter information appears to be cached and not overwritten

[flikka]

Example:

1. You built a machine with the remote_logging: True.
2. Then, you change to remote_logging: False
   If the cache gets a hit (ie. you havent changed other things in the machine), the remote_logging: change is "ignored", since the reporters patched into the machine when this was set to True were cached with that machine.

This would probably not be a big issue, BUT, we had a situation where the MLflow logging failed. But the machine was built, and cached with the MLflowReporter, so changing the remote_logging to False didn't stop the (failing) attempt to do remote logging, as the machine with the reporter attached was retrieved from the cache.

Solution: Either include the reporter in the cache key, or "patch" the machine with the information