USB Busy when transmiting data

USB doesn't have Tx and Rx, it has D+ and D-. These need to be connected by name (so D+ to D+ and D- to D-)

If you connected these the wrong way round it's possible the host will shut the port down to prevent damage. You shouldn't need a resistor from Vcc to GND anyway, the host doesn;t need to have 100mA drawn to detect a device, it detects it with a pulldown of one of the datalines)

If the hub has a separate supply, then you just need to connect the ground lines - you don't want to connect Vcc together. Obviously if it doesn't you need to power it from the host supply (i.e connect both)

The Start Of Frame signal can be used as a (slow) clock derived from the host's USB clock, which ideally should be locked to the host's general idea of time.

The focus on audio applications is because this is basically what it is useful for. If you let an audio stream run for a few hours without synchronized clocks, you are likely to have a FIFO over- or underrun, which would be audible, and deriving the ADC/DAC clocks from a host provided clock moves the responsibility to the host, allowing you to work without an elaborate resynchronization schema.

As the clock is rather slow, deriving anything faster than a few hundred kHz is going to run into stability issues on its own, so other applications don't really benefit.

What should happen with unused pins is usually specified in the datasheet or a separate pin connection guide -- if the 47k pull-up is common, it appears to be either from that document or assumed so because the last person also thought so. I'd probably cargo-cult that as well if I don't use the pin.