If I have more than one device in a signal chain with a volume/gain control, my practice has been to turn up the source to maximum (may not be 100% such as when achieving a mix between two guitar pickups), then attenuate that signal farther down the chain. My thinking is that each device is amplifying any noise already in the chain, and that the noise from low-output sources, whether line- or instrument-level is going to be the most noise-free (minimal amplification). Thus having preamps and audio amps set in a low-to-mid range with lots of headroom will provide the cleanest signal.

It gets more complicated when adding additional devices:

Instrument -> modeler -> mixer -> DAW

What strategy of level-setting should be followed in this kind of scenario?

I'm thinking that my basic strategy above is correct, but in the above scenario I should be seeking to get inputs appropriate so the gain and volume controls in the mixer/DAW are in the vicinity of 0db. So, I can leave my instrument maxed out, the modeler adjusted to get the tone sought, then adjusting the output of the modeler to keep both the gain and the sliders on the mixer in the neighborhood of 0db, then adjusting the "main out" level so that the sliders in the DAW are also hovering near 0db. Similarly, outputs from the mixer to amplified speakers or audio amps should be set so that the volume control on the amplifiers are no higher than mid-range.

Is my thinking good here or am I missing something?

Yep. That's pretty much it. Maximize the signal-to-noise ratio every step of the way.

Thanks!

Amazing that I was able to think this through correctly!