Andy Koppe reported that with both core.splitIndex and index.skipHash enabled, entries are lost or corrupted whenever the shared index is written. skipHash can be turned on via feature.manyFiles. His reproduction ends with git status aborting on a BUG() in wt-status.c about multiple renames on the same target.

Tanga Joris posted a reproduction and a cause. write_shared_index() names the shared-index file after its trailing checksum, which is written as zeros under index.skipHash. The split index then records a null base, and read_index_from() treats a null base as "no shared index", so the shared entries are lost.

Joris said this has been broken since ee1f0c242e (read-cache: add index.skipHash config option, 2023-01-06). No fix is described in the excerpts.