[mck-]

OP here: yes it is, it supports not only large-scale VRP, but also CVRPTW-type problems, which include constraints for time-windows and capacity.. Having said that, we are talking about NP-hard problems, so computational efforts increase exponentially, and it'd be a good idea to split up 100+ vehicles into smaller chunks.

Unless your deliveries are done overnight, it's no problem to run the algorithm through the night.. It is capable to find a quick solution (~80% optimality) in very little time, but is able to keep searching for as long as you want to improve on the solution.