stageUnorderedWritesToBucketCatalog and stageUnorderedWritesToBucketCatalogWithRetries take in a startIndex, and take it into account for some error handling, but because prepareInsertsToBuckets does not take in a startIndex it does not actually get respected.

This results in an entire batch getting staged again even though we may only want to stage a subset of it, such as the fallback case from ordered writes, where we want to retry each measurement one at a time.