Solver failures with source strengths in all-to-one

[bmcrae]

This is the error reported by Catherine on May 5 2014. All or almost all solves fail under all-to-one mode when variable source strengths are used. Input files that replicate the error are here:
https://dl.dropboxusercontent.com/u/1340127/CS/mailCS_div1.zip

[bmcrae]

Note that the example problem had variable source strengths from 10-170,000, in floating point. When Catherine reduced the range by log transforming these, the error went away. So this may be a bona fide solver failure and not a bug.