Great question. Generally our system performs better in higher volume areas given the higher "n" in the data but we don't have a live deployment in a high traffic community ER yet. Hopefully we will have a more conclusive answer for you soon!
Does it get reinforced for doing an obviously bad or good job?
Yes. Our data feeds include the outcome variables we are optimizing for which helps the algorithms retrain as new patterns emerge.