The regression tests in gs/toolbin/tests should be updated to run ray's cmpi tool in batch mode to compare the two rasters and report the results. Additionally, if the differences are trivial as defined by the cmpi tool, run_regression (or probably better, run_nightly) should update the baseline automatically. Having a human look at every small change isn't worthwhile, and letting such changes build up over time shadows new differences.
This should follow the manual review of the current regressions just so we know that the cmpi method seems reasonable and which params are best at 72 and 300
This may not be an issue anymore since we haven't been tending to build up regressions as much, but we have discussed using fuzzy/cmpi to "filter" the results from regressions (including on the cluster) Assigning to Marcos, but feel free to discuss closing this.
I'm closing this. We have so few trivial nightly regressions that it's not an issue to examine the files manually (even if we did I'd be concerned about missing stuff so would end up examining all the differences anyway).