I am using the Ozone 8 'Maximizer' in my audio output stream "full-time". I use "Audio Hijack" to take the audio and pass it through my room correction software. The normal path is
Audio -> O8 Maximizer -> ARC -> output device
I use the Maximizer to dither the output - my primary interface is 16-bit/48 KHz - threshold and ceiling are at 0. I honestly can't tell the difference between 24 bit/48 KHz and the 16 bit version dither or not.
In all of my mixes I have O8 last on the mixbus (I don't dither here) and I do preview codecs when mixing music for YouTube videos.
I can spare the 50% of 1 cpu thread that is consumed by my audio output.