That works!
I think I understand all the individual bits of what you did there well enough to implement it but the syntax is is going to take me a bit longer to unwrap.
The solution doesn’t have to be unique for what I’m doing, but I’ve found that I could herd GRG solver in the right direction by searching for intermediate solutions. That makes it less likely “cheat” to zeroing out some of the weights in my implementation. Perhaps something like that will work here too.
In any event: Thanks to both of you again.