Currently, the percent suboptimality graph is quite slow if not cached. This is because naively, we must join the instances table with the submitted instances to obtain the current best lower bound. We can make this better but it takes a bit of work to modify the querying system for this.