We address the 'rich' (i.e., with real-world features and constraints) inventory routing problem for bulk gas distribution under uncertainty. We consider that the uncertainty occurs on the supply side and consists of outages at the production plant. We propose a general methodology for generating, classifying and selecting 'robust' solutions: solutions that are less impacted when uncertain events occur such asplant outages. This methodology is applied to real data provided by the Air Liquide company in the context of bulk gas distribution, and we show that for a relatively small increase in cost, the robustness of routes and schedules for the bulk gas distribution with regard to possible plant outages is improved. Results show that we can reduce the extra cost induced by plant outage, while only slightlyincreasing the cost in the cases where no outages occur.