In terms of potential explanatory covariates, I think preferred editor/IDE would be one of the most likely to explain the trend, since different editors will ha

In terms of potential explanatory covariates, I think preferred editor/IDE would be one of the most likely to explain the trend, since different editors will have different defaults. The survey has this information, so someone could test this. It would probably help to group the editors by their default setting. I think you might need a mixed model with editor as a random effect to include both default setting and preferred editor in the same model.