I just did an experiment to see if I could play a DVD and a YouTube video at the same time. To my surprise, I could. The Focusrite showed its sample rate as 44.1 KHz (my normal rate for everything, and the Windows default). Given that the DVD audio is 48 KHz, that tells me Windows is quietly doing a sample rate conversion behind the scenes. I'm guessing that if I had designated 48 KHz as the default, Windows would have done the conversion for the YouTube video instead of the DVD and all would still be well.
Note that Windows explicitly labels this setting as "the sample rate and bit depth to be used when running in
shared mode", an acknowledgement of the fact that
any interface can only handle one SR and bit depth at a time.
So back to the OP's question: why doesn't this always work? My guess is it's because of ASIO, which bypasses much of the Windows audio subsystem and therefore cannot benefit from this automatic SRC.