A DS+RS starts to run low on disk space. We log the "low disk space" message and reject local changes with unwilling_to_perform, but we continue to accept and write changes into the changelog, and replay them in the backend.
However when the disk becomes full, writes to the userRoot fail with an IOException (no disk space) and the userRoot environment is invalid.
But the server keeps running, and seems to keep trying to write to the changelog. This is probably not sensible, and we should consider preventing all DB writes (including JE cleaner threads) before this happens. Perhaps shutting down the server?
If we keep running we can end up with a corrupted backend, and a changelog that can't be opened. (Even if you stop the server, grow the underlying disk, and restart the server.)