Due to reference count juggling, we can easily end up in a situation where an instance of a subclass of GProxyVolumeMonitor is disposed but not yet finalized. For example, this happens in situations where you run 'gvfs-mount -d /dev/sdb1'.

I discovered this problem while developing an udisks2-based volume monitor (to replace the gdu volume monitor).

While it's possible to fix this [1], a much better solution is just to keep the remote volume monitor proxy around forever.

[1] : for example, by moving

 g_hash_table_remove (the_volume_monitors, (gpointer) type);

into the dispose handler... but this fix alone is not sufficient.