2) Dragon 2 demonstrates on-orbit operation with ECLSS/consumables for the length of the free return flight.
I could see a failure of such a demonstration pose some serious questions with regard to NASA's certification process.
You misunderstand the point of such testing.
Testing always delivers confidence. Especially if it fails, because you bring on such a failure
in the context of well-planned and on the scene contingencies. You won't have these during actual missions.
Always think "belt and suspenders". As to certification process, it either finds holes for Starliner/Orion, or it confirms the process with actual flight data. There's nothing bad. And its cheaper than EFT-2.
Now, for NASA only - such a capability for long duration flight is a substantial proof for a potential LON capability, extremely cheaply obtained. What else would you need:
1) additional delta-v to allow Dragon-2 to enter/leave LLO (or other). (Or only fly Orion to near lunar halo/EML orbits both can already reach.) Possibly extended tanks in trunk, or propulsion pallet in trunk.
2) demonstrated unmanned mission (Dragon 1) to validate heat shield. Something akin to ETF-1, might even be possible with just a F9 on a highly elliptical orbit.
LON mission concept: Before flight, prep a FH/Dragon-2 for flight with propulsion pallet in HIF. On emergency, prepare/validate for flight, erect, launch.
If no emergency, remove propulsion pallet from free return commercial flight that would fly in a week.
Now, its likely that Orion flights would never need this. However, the public is very unforgiving of "Marooned" astro's, out there suffocating. And at the moment, Dragon-2 would be the "best bet". Cheap contingency.
Don't you think?