As promised I am asking for help to write a regression test for this bug.

First, I want to make sure that time spent in EA does not exceed EscapeAnalysisTimeout. Also this is new product flag and needs to be tested.

Second I need a stress test for EA. Next class which implements BTree shows the problem - a lot of array allocations and arraycopy (Arrays.copyOf()) in compiled method (after inlining):

https://github.com/karussell/graphhopper/blob/master/core/src/main/java/com/graphhopper/coll/GHLongIntBTree.java

Also it is not clear what our interaction process is.
Should I file a separate bug for new test or I should wait when the test is ready and integrate it together with the fix?

Note, I reproduced the failure using 7u80 sources and steps described in 8041984 bug report. After building graphhopper binaries you can run standalone java application and observe long compilation time.