Metrics reporting too often

Barnes was re-written to report to the Heroku API directly instead of statsd -> agentmon -> Heroku API. In this process, I think the reporting cadence (polling loop) changed so we're now sending data more frequently to the backend. I need to investigate.

No file or test is named. Trace Barnes's polling loop and compare the current direct Heroku API reporting cadence with the previous statsd → agentmon → Heroku API path; done means identifying why reports are more frequent and verifying the expected cadence.

Written by the indexing model from the issue text.