Describe the bug
In #9333 a new Feature got introduced called MaxFrontierSize, which limits memory usage in case of high density graphs. This causes a algorithmic issue, because the "best" node is droped from the priority queue when the MaxFrontierSize is triggered.
This causes a nearly 100% to not get the shortest path ever.
To Reproduce
Steps to reproduce the behavior:
- run shortest path query with numpath = 1 to get the shortest path
- run the same query with numpath > 1 with MaxFrontierSize low enough to get triggered on your data set
- Compare Paths
Expected behavior
numpaths > 1 should return the somewhat shortest paths, possibly even the correct shortest paths