DisplayCoreCustomCompositor errors

I have some HMDs that I've added the special EDID extension to mark them as non-WinMR HMDs, and I'm trying to use the sample provided in this repo. They are correctly recognized, but are triggering an hresult_error at one of a few places:

• The first time I run it, it throws here, on the actual call to TryAcquireTargetsAndReadCurrentState: https://github.com/microsoft/Windows-classic-samples/blob/340d6c8c537919663cb122d1a3cf76bed91c06f8/Samples/DisplayCoreCustomCompositor/cpp/DisplayCoreCustomCompositor.cpp#L283-L285
• On subsequent executions, it throws a bit earlier, here (the preceding state TryApply): https://github.com/microsoft/Windows-classic-samples/blob/340d6c8c537919663cb122d1a3cf76bed91c06f8/Samples/DisplayCoreCustomCompositor/cpp/DisplayCoreCustomCompositor.cpp#L279-L281

In both cases, the hresult returned is "0xd00000f2" which is a complete dead-end when searching, and no further message is available through WinRT.

My system is running Windows 10 1909, on a Radeon RX580 graphics card (latest WHQL driver, Driver Version
19.50.29.27-200515a-355311C-RadeonSoftwareAdrenalin2020 ). The HMDs I'm trying are two modified OSVR HDK headsets (one v1.3 without HDCP, one HDK2 with HDCP), both with a modified EDID to add the non-WinMR HMD flag. Also tested on a Dell XPS13 9370 (Intel graphics) - same result for the HDK2, it can't see the HDK1.3 at all (which I think I expected). Also tested on an NVIDIA GTX1050 - same result for the HDK1.3, "no matching modes" for the HDK2 (which seems unrelated).

I'm currently updating the NVIDIA machine to test on 2004.

Start with Samples/DisplayCoreCustomCompositor/cpp/DisplayCoreCustomCompositor.cpp, especially the TryApply call at lines 279-281 and TryAcquireTargetsAndReadCurrentState at lines 283-285. Run the sample with the reported HMD and inspect the HRESULT handling around both calls. Done means identifying why HRESULT 0xd00000f2 occurs and documenting or fixing behavior so the sample can proceed with the affected headset.