Intersecting and fracturing zone fills takes forever for complex
fills (VCU118 has some fills severa with hundreds of thousands of
points).
However, they are trivially parallelisable - so do that and cut VCU
board load times by 70% - YMMV depending on CPU.
Further save 10% by sorting the heaviest zones first which is nice,
but the real thing to avoid is accidentally scheduling the biggest
zones consecutively on the same thread, which could be a substantial
penalty.
This has the happy effect of reducing the Allegro test suite to under
a minute (VCU118 dominates)
John Beard