The schematic with Q10 is exactly the same as the first post, although the lower side of the bridge now matches: I've added "Q11" between Q8 and R8. (which doesn't affect measurements either way).

I'll get a new meter tomorrow. Thanks for the legwork, but I've actually had the manual sitting by my keyboard since I went to look up the model name

A couple of updates:
I realize I must have damaged 2 of the transistors the first time I tested with the current mirror and a 1M bias resistor. Not being proficient enough to know whether the others were okay, I set them all aside and rebuilt the circuit with new transistors.

More testing has shown that tripling R7 and R8 from 10k to 33k did NOT solve my \(\strike{voltage}\) current problem... it only made the problem less apparent.
The current across the load (my meter) only remains at ~2.6mA for about ten seconds, and then slowly begins an exponential climb up to frying transistors at .11A.
Metering the current between Q10 and R7 showed that the current remains constant on that side of the mirror while the current across the load side is slowly climbing.

All the iterative testing is surely helping to drive some principle into my head (once I figure out which that is), and reading resistor color-codes now comes intuitively... but I'm hanging up my hat until tomorrow when I have a more appropriate meter.

night night,
-masked

 

Voltage is inmaterial in the constant current scheme of things. Now if you were measuring the voltage across the known loads and calculating the currents ... that is a different story, and one we could have investigated with your ideal 360. The currents should have been closer to the 2.6 mA you wanted. This is not to talk you out of getting yourself a better meter for the various measurements.

I would get back to Ron's design ... as it is a lower component count.

 

I guess I fried my brain along with that transistor.... it made me type "voltage" when I was thinking "current".

Sorry if it looked like I was ignoring his considerable help. I agree on Ron's design; His 8 transistors / 6 resistors is simplest.
My 9 trans. / 6 res. was an experiment in adding 1 component to "save" amps. I even added one more transistor when I wanted to test each half of the bridge independently, rather than using a single transistor to regulate BOTH sides of the bridge as Ron's model ingeniously does.

I'll go his route eventually (with a wall wart too) but, as this is my first project in 20 years, I'm trying to understand every variation of this known circuit. If I just copy the first schematic I see, I'll stop playing and stop learning.

-masked