Just to give a personal anecdote: in 2015 I was working on large scale vehicle routing applications, where I used JGraphT to represent entire road networks of some major US cities. The data came from Open Street Maps. The graph representation was simple: a road intersection was represented as a vertex, an arc represented a directed connection between a pair of intersections. Each arc stored information such as distance, shape, travel speed, number of lanes etc. Within this large road network I had about 3000 Points of Interest (POIs). These POIs formed a small subset of the vertices. For my application I had to compute all-pairs shortest paths between my POIs using the underlying road network. Back in 2015, JGraphT had limited support for routing algorithms. With a modified version of Johnsons shortest path algorithm I was able to compute the desired distance matrix in 30-60 minutes. Recently I had to revisit this application. For a food delivery application for students affected by COVID-19 I had to compute a shortest path distance matrix for 3500 POIs (the students) in a road network graph with 35296 vertices, 84887 edges. Using the Contraction Hierarchy algorithm (CHManyToManyShortestPaths) implemented by Semen in 2019, I was able to solve this problem in just 15 seconds! So the same application that took 30-60 min in 2015, is now solved in 15 seconds. Personally I'm very impressed by the talented developers that have contributed to JGraphT over the years, as well as the level of maturity that this library has reached.
We've been receiving a lot of significant contributions from Semen Chudakov over the last year and a half, so we're very happy to announce that we're adding him as a committer on the project (joining me, Joris, Dimitrios, and Timofey). This means if you submit a pull request, he may be one of your reviewers helping you get it ready for merge.