The old behaviour was such that if none of the normal mappings worked, we would probe ALL fallbacks. I don't think that makes sense, and it caused concrete issues: let's say we have a regular stereo mic device, but there's no "front" PCM defined for it. In this situation we would probe the stereo-fallback mapping (which uses "hw" instead of "front"), and it would work, but then we'd also probe the "multichannel-input" mapping, which would also work, so we end up with two mappings that don't have any difference in behaviour.
I think it's better to simply pick the first working fallback and ignore the rest.
Fixes: https://gitlab.freedesktop.org/pulseaudio/pulseaudio/-/issues/901 (issue is marked as confidential due to unreleased hardware)