Fargo doesn't watch NameOwnerChanged to know whether the CM has crashed, and doesn't seem to terminate calls if the CM spontaneously disconnects (e.g. network error) either.
This turns out to cause Bug #26359: when a media Channel object path is re-used, the old callbacks are still attached, Fargo tries to do everything twice and failure ensues. Upgrading priority and claiming this bug.
*** Bug 26359 has been marked as a duplicate of this bug. ***
Estimate reduced to 3 hours, this went quicker than I expected.
Branch smcv/misc is ready for review, but I haven't verified that it fixes the symptom from Bug #26359.