The newer capture methods are great ideas, but they still have bugs, and possibly unfixable problems if the operating system itself doesn't even have the concept of what OBS wants from it.
The older methods of capturing *whatever* is on a particular screen or comes out of a particular sound card, are practically bulletproof, but you then have to control by other means what actually goes there. If you just capture your normal headphones, and you use those same headphones to Monitor yourself or the app, you'll get an echo because it really does capture *everything* that goes there.
---
You might get a cheap USB sound card and not wire it to anything. Just plug it in somewhere, send the app(s) to it that you want OBS to capture, and nothing else.
Or you might install a loopback audio driver. That'll appear as a new speaker and a new microphone. Send the app(s) to that virtual speaker, and have OBS pick up that virtual mic.
VB-Audio Virtual Cable and App's
vb-audio.com
Or, you could even have a mixing console in between:
VoiceMeeter Banana, the Advanced Virtual Audio Mixer by V.Burel
vb-audio.com
VoiceMeeter Potato, the Ultimate Virtual Audio Mixer for Windows
vb-audio.com
Two different sizes of the same thing. Each of them installs some virtual speakers and mics, that connect to the virtual mixer. It's as if you had a basic physical mixer and that many physical sound cards plugged into it. Connect OBS and your other apps to the appropriate ones for what you're doing, and use the mixer to complete the connections.