[r-t] Big searches

[Mark Davies]

DFM:

> Were that to happen with something I was working on I believe my thinking
> would go something like "That's surprising. I suppose it *might* be that
> the order in which my search is doing things happened to go down the best
> route early, but it does seem unlikely. I wonder if maybe it's a bug in my
> code?

Ah, but the software's been in beta since 1998... surely there can't be any
bugs left... ;-)

Speaking seriously, I think it's reasonably easy to explain. The search
starts from a Before in the plain course, and a series of Befores after that
rates up pretty highly on the music and avoiding-falseness scales. (We're in
a cps method but there are non-searched split-tenors blocks being linked
in which knock out a good few nodes, effectively meaning some parts of the
truth table are marked as false before the search starts).

MBD