Funny you should ask about a tree resting on a phase. Under the right conditions that runs about 400 ohms of fault resistance. The current in the faulted phase at the sending end goes up, a bit, and the current in the faulted phase at the receiving end goes down a bit (but doesn't reverse direction). The POTT scheme got it without a hint of hesitation.
We use both phase and ground distance elements for stepped distance and avoid overcurrents for normal, non-comm, tripping. But for the POTT scheme we use the phase zone 2 distance and a directional ground overcurrent. Since the ground overcurrent only has to securely get direction correct we don't have to worry about reach and can use some fairly low settings. Using SEL relays, as we do, everything since the 321 has included a reverse block and some echo logic. It is essential that the reverse element at each end is more sensitive than the forward element at the other end, in primary current. But other than that just stay away from the zero-sequence polarization if there's any possibility of mutual coupling.
Differential doesn't hurt, but in my experience the POTT is never slower and often faster than the differential. Throw a DUTT on top of it, and you'll find that the DUTT/POTT combination is almost always faster than 87L on a clean two-terminal line. The extra comm time and data processing can leave the 87L slower by a quarter to a half cycle.
Turn that into a three-terminal line and the 87L becomes more important. The newest line differential we work with includes traveling wave fault location; that's worth it all by itself, even if the DUTT/POTT always beats the 87L.