It has come up a couple of times that it would be good for the SDK to support cursor-based syncing. For APIs that allow it, this would allow resources to be fetched once, with only updates being fetched after the first sync. For some resource types, this optimization would be a significant performance boost.

This would be especially useful in cases where the APIs allow for time-based querying, and it is known that resources never change once they are created (like logs, reports, audit trails).

One concrete example where this would be beneficial is the Heroku audit trail event: https://devcenter.heroku.com/articles/platform-api-reference#audit-trail-event . It allows querying by day, so an optimized implementation would only fetch historical days the first time, then in subsequent runs fetch the days that have been added since the previous sync.