The fix for JDK-8134953 tried to keep creation of new GC ids the same as before. But there was one exception. When we create heap dumps or do heap inspection before or after a GC we no longer create a new GC id for those events. Instead they use the same GC id as the "enclosing" GC.

This was an intentional change. But in retrospect it should probably not have been done as part of the other changes.

Here's the change to collectedHeap.cpp that removes the creation of new GC ids:
http://hg.openjdk.java.net/jdk9/hs-rt/hotspot/diff/983c56341c80/src/share/vm/gc/shared/collectedHeap.cpp

This breaks some tests that expect the GC id for the heap inspections to be different than the GC id for the actual GC.