Andy Koppe reported that with both core.splitIndex and index.skipHash enabled, index entries are silently lost or corrupted whenever the shared index is written. The second setting can be turned on through feature.manyFiles. In his reproduction, git ls-files printed two bare "./" lines, and git status --short aborted with "BUG: wt-status.c:585: multiple renames on the same target? how?".
He said the problem appears without an explicit git update-index --split-index once enough entries change that the shared index is rewritten. It can also appear in a new repository, where the first git add writes a shared index.
Tanga Joris reproduced both cases on master (fb255012ac) and gave a diagnosis. write_shared_index() names the file after the shared index's trailing checksum. With index.skipHash that checksum is written as zeros, so the file ends up named sharedindex.0000000000000000000000000000000000000000.
The entries disappear because the split index then records a null base. read_index_from() treats a null base as meaning there is no shared index (read-cache.c:2381), so the entries stored in the shared index are lost.
Joris said he is working on a fix and will send it. Koppe thanked him and said he had not dared to wade into the code himself. No patch has been posted yet.