Some time ago had to create a workaround in connman so that BT rfkill initialization during bootup works predictably (other alternatives were tried, but were found to be more difficult to implement reliably).

Due to weirdness in connman code the presence of a dummy BT rfkill driver that doesn't have tethering functionality prevents tethering by the "real" driver; not difficult to change the code in the problematic place but should understand why it was written as it is in the first place.

Problem is that while technology.c:set_tethering() should iterate over drivers once it finds one that can attempt tethering, technology->tech_drivers isn't updated when drivers are added if the technology struct has already been created.

See set_tethering() at https://github.com/mer-packages/connman/blob/master/connman/src/technology.c#L234, technology_get() at https://github.com/mer-packages/connman/blob/master/connman/src/technology.c#L1168 and technology_driver_register() at https://github.com/mer-packages/connman/blob/master/connman/src/technology.c#L1249