Some errors present in parallel programs can be very difficult to detect during testing as they may only occur under special circumstances and may not be repeatable. While Analyst Drive has been extensively tested and should be considered rather stable in a parallel setting, it is possible that certain parallel errors still exist and just haven't surfaced under the right conditions. Should the user find that a solution is much different than expected when run in parallel, they should save all available files produced during the problem run and then re-run the program setting NCORES and MPICORES to 1 to see if a much different solution is achieved. Should the results indicate a solution extending beyond reasonable error bounds as described above, then this would indicate a parallel coding error that should be reported to Citilabs support (generalsupport@citilabs.com).