This bypass transistor does not have to take the full current. It seems that Q1 kicks in after input current generates big enough voltage drop over R1. So for a certain current range, Q1 could act as a parallel current source and power is shared between linear regulator and the BJT.I wouldn't do that. Your bypass transistor still has to dissipate 7W, or more like 7V @1.5A = 11W
What are you using as your power source and what are you supplying with 5V @1.5A?
If you reduced the input supply to 7.5V and use a LDO regulator, you will be dissipating 1.5A @ 2.5V = 4W.
You will still need a heat sink though a smaller one.
Challenge for me is to calculate the proper R1 value.
And you have a good point, I'll see if can use lower voltage than 12V for the rest of the circuit.
My plan for now is to use LM340S in DPAK (TO-263) package and add the ability to include this bypass transistor when heat becomes an issue (add heat sink there). Also do a lot of vias to ground plane (for LM340S) and such fun.
But this is getting far away from the original topic.. should I start a new thread if want to continue this discussion?