writer node: smarter healthcheck

[roadscape]

During sync, ideally we'd be checking rate of progress at each stage.

We may need a separate lightweight db to store current state so that while postgres goes down we have access to status and last 'checkpoint'. The simplest implementation could update a last checkpoint date + rate of progress. The process is healthy if the last checkpoint was within last 30 mins and rate of progress meets some reasonable thresholds.