If the same commit is benchmarked more than once, it leads to duplicate points on the graph


I think it should be possible to dedup these entries (primarily caused by manual runs) in the client-side JS (unless there is a way to prevent this in the GitHub action that publishes these records).

Start by locating the client-side JavaScript that loads benchmark records and renders the graph, then inspect how manually published records are represented. Confirm whether duplicate entries can be identified by commit, and consider the GitHub Action path only if the client-side data cannot be deduplicated. Done means repeated benchmarks for one commit produce a single graph point.

Written by the indexing model from the issue text.