But most importantly, I agree with your very last sentence, we're all different and what's optimal differs a lot from person to person.
I have folders full of sketches I've built up in years of using it but very few things I'd call finished.
From there, I see it the way a sculptor might see a slab of rock, and start “carving out” the song.
The OP-1 is where I've found this to be especially the case, incredibly easy to make a loop on but never managed to make a full song on it, aside from using mutes/effects on a single loop to stretch it out into a full track
But that initial barrier is such that a beginner will be able to get much more mileage out of such a device. They might never even run into the limitations though it may limit the musical space that they can work in on some dimensions.
In the end what matters is the result, not so much how it was reached.
I grew up playing classical guitar and use to own a Ztar and this was the issue I had with it. My guitar chops transferred to it pretty nicely but it is still its own instrument. Own instrument with no music or practice material, no transcriptions of anything other than what you come up with yourself.
How far would anyone get learning piano if they had to also come up with the music and practice material outside of drilling scales concurrently with learning the instrument itself? It would just be a nearly useless gimmicky instrument.
As with anything, it does just take practice, and these instruments are so much easier to play than something like a guitar that just noodling around will also get you places.
Sunvox is pretty great for quickly roughing out an idea. I use my MIDI keyboard for building patterns. I'll record the notes for a pattern and then play it back in a loop to adjust the rest of the parameters. I use the other mapped controls live until I have what I want. I prefer having tons of physical controls ready to go for this part. I have a few project files I always start from and their corresponding templates are saved into the keyboard. Then I go into audacity and record my guitar, voice, or whatever else to a click and import them into the timeline. I think it's more important to have your workflow down than be distracted by new shiny toys.
But one of the main things that saves time is how loud or soft each note is. Doing this in a DAW is far more tedious, neglected completely for just repetitive electronic music, or requires scripting to provide random-ish entropy to each note. Whereas with a Push its based on how hard you press.
To add some objectivity to this, I would say its more about the learning curve, I couldn't expect someone to pick up a Push and get anywhere fast with it.
But I think people familiar with a keyboard and mouse would benefit from a Push and accelerate their work.
I prefer to be DAWless when I am in "discovery mode" or just want to jam around like with a regular instrument. However when I am ready to assembly everything into a song I prefer to use a DAW.