_scroll comes with limitations such as the fact that it only allows to go forward, which can be an issue if there are connectivity issues while the client is retrieving one page as this particular page cannot be retried.

We would like to replace _scroll with an API that allows to open point-in-time views of the data and then allows to paginate through it using search_after. We could even keep the sort by _doc optimization if users provide "track_total_hits": false in their request.

In FixitFriday it was suggested that we call this API Jim's API.