Maybe I missed a known bug when doing the research. But all the other reports describe different and reproducible behaviour.

Sometimes (often) I cannot unfold a subtree. It then affects only one or more subtrees while others in the same buffer or even the same branch still work. The only fix then is to either cut and paste the affected branches, or to close and reopen the file.

While quite frequent I have not achieved to reproduce the issue. It just happens after a while of using multiple files navigating with search, tab from agenda, org-refile, and cycle.

Based on my analyses it's not the cycle functions, but must come from broken text attributes. Even (org-fold-core-region (point-min) (point-max) nil) does not unfold the affected subtrees then. Assuming that (org-fold-core-region) is the only place where the folding and hiding happens, I expect the problem there. But I understand too little about the emacs internals to debug this function.

I hope this helps to fix the issue.