I think this bug report may be about the same issue that I came across a while back here:
https://www.spinics.net/lists/git/msg316628.html I found that the problem was "introduced" in d0c39a49ccb5dfe7feba4325c3374d99ab123c59, first released in 2.15.0 (as also noted in the bug). > revision.c: --all adds HEAD from all worktrees > > Unless single_worktree is set, --all now adds HEAD from all worktrees. The problem shows up if you have a git worktree where the HEAD revision ends up being expired (which happens automatically). My understanding is that the problem is that git prior to 2.15 could corrupt your repo by pruning a still-accessible worktree; this no longer happens. I haven't actually verified this myself, but since I sorted out my original broken tree, and moved to 2.15, I haven't seen the problem again. I think you could verify this theory by creating a worktree, expiring it and then doing fsck, with a 2.14 git and a 2.15+ git, but I haven't done that myself.
