There are a few other similar questions here (search for “Too long with no output” to find them).
I posted on one recently. I had wondered if their process has spawned processes that CircleCI was waiting to exit, but it seemed not to be the case.
Are you able to add logging to see exactly where it is getting stuck? I wonder if a trace tool, applied via an SSH console, might help? Perhaps that would reveal what kind of operation it is waiting for.