Stereo Tool

Hello everyone, I want to ask you a thing!
Scenario: small university webradio. Music -> ADC -> mixer (with voices) -> DAC -> Stereo Tool.
In the present we usually import CDs normalizing in peak (-13 dBFS), therefore saving some headroom. By doing this what happens is that high-dynamic old rock songs play quieter compared with compressed modern rock songs. As we do not gain in the mixer stage, we can say that when music gets to ST it is levelled in peaks but not in loudness. ST does its job (AGC) and produces a homogeneous sound. So far so good.
But what if we normalized in loudness *before* music goes into the DAC? I was thinking about taking everything down to (just to say a value) -15 LUFS, so that all the catalog has same loudness and different peaks. Provided that nothing clips in the analog (mixer) stage, it will reach Stereo Tool with the same characteristics, same loudness, but different peaks.

From Stereo Tool's point of view would this second solution be better compared to our present configuration?
My idea is that by doing this AGC will work less and some songs (oldies) wouldn't need to be lowered when it is not necessary.
What is your opinion?

Thank you!

That's definitely better. The AGC should be able to handle almost anything you throw at it but making things easier for it never hurts.

Nice to hear that, thank you!
And in your opinion what would it be a reasonable input level for Stereo Tool? Maybe it is a well discussed topic in the forum, in case you can address me to some previous threads...
I suspect Stereo Tool would need some headroom in order to declip or restore some lost dynamic to the incoming "squared" music, but how much? Are we talking about 1-5 dBFS or 10-15 dBFS?

It doesn't need any headroom (you can input the audio at 0 dB without any issues) but you have to make sure that your most dynamic track can still be normalized to the target level without clipping.