That would be exceptionally sloppy development. Phones have had more than enough power for long enough. 4 core Skylake (Mac 2016) would be well beyond human capabilities, if it's just raw power.
The "thinking" (difficult) limit should be considered moves ahead, both depth and count. With a possible limit to time, if there is any time control.
IIRC it does just set a time limit on thinking
Not if the computer's time limit is set at 15 microseconds. It's not a question of whether the computers have "enough power"; just whether they are more powerful now than they were previously.
And yes, obviously that's a very sloppy and error-prone way to implement a difficulty control.
Carlsen knows how to play anti-bot chess where some engines may struggle, but that only applies to amateurish engines.
I fired up Chess shortly after getting an M1 and got destroyed a bunch of times. I thought that I was just extremely out of practice and quit playing for years. I guess it's better to find out late rather than never.
Heck; even Nanochess was rough for a novice like me, and that on an n270 CPU.
In that case you'll hit issues on any device that performs significantly differently from that which it was tuned in.
Though I am slightly amused by people using the apple chip as an example of "high performance" in a problem that scales very well with threading.