There have been some reports of "plate scale too far off" failures when really it's not that far off. See this Comm Center thread by Jesse Goldblum. The crazy thing is that if you reduce the plate scale from the exact 2.07 value to just 2.0, it succeeds, and will continue to succeed all the way down to 1.7, or 30% low. It looks like the acceptance band is 15% low! How can this be?
This is probably some sort of edge case and really needs looking into! I allocated 8 hours for the inevitable regression tests even if I find it quickly.
See attachments for test setup.
