Metric VNode Repair window

Created by: davecromberge

@Licenser I would like to clarify my understanding of the boundary checking within the handling of the repair command within the Metric VNode.
I would like to use illustrations along the way, hopefully this will help in articulating my point. All cases assume that the cache contains values for the given Bucket and Metric pair.

I feel that there is overlap between conditions 2 and 3, and would like to propose the following amended condition 2 (in accordance with my understanding):

I assume that the number of points residing in the ETS cache does not have to be the same as the number of points in a repair, seeing as a repair could be propagated in response to a range query (with variable size). Does clearing the cache before flushing to disk increase the chances of data loss, if writing to the file system fails.

I would be grateful if you could also address what you refer to as an intersection for the cache and repair data. Are you referring to the case where at least one value overlaps?

Thanks.