In my understanding, setting a higher sample rate will cause the OS to digitally upsample or downsample the source material to match. I don’t know if it’s audible from an ABX test perspective, but it mucks with the signal. I suspect it doesn’t matter for upsamples of multiples of the original (like 44kHz to 88kHz or 48 to 96), but interpolation (44 to 96) might introduce artifacts. (I might be wrong about this.)
A higher output bit depth than the source material should be fine: the extra bits will be set to 0, and that’s that. No information added, removed, or changed.