LTR tree traversal currently happens in a recursive way.
Background :When we tried to increase the number of trees in our production solr, we were facing latency issue. From solr profiling it came out that most time is spent in recursive call for tree traversal.
Recursion is slower due to overhead of function call stack, so we optimised this and translated it to loop based approach. This improved solr latency significantly and we were able to scale much better.