question about min hash deduplication process, how many .remove files I should be expecting with 14 buckets?

for minhash deduplication with 14 buckets after running stage3, I only get 2 .remove files. I see that we are only using 1 worker, so I think this is expected to not be 1 .remove per bucket?

when running stage4, I see message complaining about no .remove found, but the pipeline continues with multiple localpipeline worker nodes with offset configured.

Thanks

Start by locating the MinHash deduplication stage3 and stage4 entry points, then trace how 14 buckets, one worker, and offset-configured localpipeline workers determine .remove files. Reproduce the run and compare the stage4 no-.remove warning with the files produced by stage3; done means the expected behavior and any required change are clearly established.

Written by the indexing model from the issue text.