Well, doing some looking around and some tinkering, I thought I'd throw a 2k2 resistor (or something small in that range, feel free to tinker) between the 102/103 and the supply after the trim. The result? A tiny bit of a more treble response, maybe the tiniest dip in gain (oh hey, there's a nice trimpot there to fix that) and substantially less noise.
Why this wasn't done to begin with, I don't know and admittedly I never spotted it many years ago when I was reverse engineering it. I even reopened the original to see if I had missed something and nope, even the original had it omitted.
Basically, these are the articles I have to thank.
https://sound-au.com/articles/cmpd-vs-darl.htm
https://www.electronics-notes.com/artic ... y-pair.php
The bypass resistor is included to aid the switch-off process. Without the resistor in place, there is no discharge path for the base emitter capacitance. Including it enables the charge stored in this capacitor to dissipate and this aids a faster turn-off.
It is good design to include this resistor, but if speed is not an issue then the resistor can be omitted, but unless cost and component count are being reduced to the minimum then it is wise to include it.