When testing the impact of some sources on a model perimeter, I noticed a bug where they were assigned the wrong TRT. Correcting this, with no other changes, led to odd results: some sites that are well outside the integration distance (e.g. >600 km where it's 300 km max dist) of the impacted sources are changing hazard by >0.05 g which is sometimes >20% and therefore non-negligible. These calculations use full path enumeration so it cannot be a matter of sampling. In the log, I do see a difference in the number PointSource ruptures and CollapsedPointSource ruptures reported in the log. I will share a test privately.
When testing the impact of some sources on a model perimeter, I noticed a bug where they were assigned the wrong TRT. Correcting this, with no other changes, led to odd results: some sites that are well outside the integration distance (e.g. >600 km where it's 300 km max dist) of the impacted sources are changing hazard by >0.05 g which is sometimes >20% and therefore non-negligible. These calculations use full path enumeration so it cannot be a matter of sampling. In the log, I do see a difference in the number PointSource ruptures and CollapsedPointSource ruptures reported in the log. I will share a test privately.