This is pertaining to the scroll API. I observed when scrolling and encountering an error, (In my case was a bad gateway , 502) , I will never get through the full population.
This observation tell me that, my request did reach elasticsearch and my scroll was forwarded to the next frame/page. Thus, I feel that as part of the scrolling, in addition to the scroll ID, the total frames/pages and the return frame id should be included. This will allow users to handle such skip.