After introducing support for multiple large page sizes in JDK-8256155 the function os::large_page_size() is a bit ambiguous. Or not really, because it will return the largest large page size available for use. Even so it might make sense to look at not relying on os::large_page_size() and instead always the _page_sizes "bitmap" containing all configured page sizes and helper functions like page_size_for_region*() to figure out which page size to use.

Currently this would probably be quite hard since a lot of the heap setup is using os::large_page_size(), but when JDK-8267475 gets addressed we could look at doing something like this.