Yeah, the author is not arguing against using 24 bits when recording, just when distributing to end users.
If the producer is planning to slow down the audio (and wants the ultrasonic components to become audible), then recording at higher sample rates makes sense, and the author doesn't address this; probably this is pretty rare in practice. You'd also need ultrasonic-capable microphones.
The much more common operation is to filter or amplify the signal, and for that, more bits per sample is better to avoid amplifying your quantization error. The author covers this in the "When does 24 bit matter?" section.
If the producer is planning to slow down the audio (and wants the ultrasonic components to become audible), then recording at higher sample rates makes sense, and the author doesn't address this; probably this is pretty rare in practice. You'd also need ultrasonic-capable microphones.
The much more common operation is to filter or amplify the signal, and for that, more bits per sample is better to avoid amplifying your quantization error. The author covers this in the "When does 24 bit matter?" section.