This is an attempt to realize Pavel's suggestion from D48393
It's even more complicated than that, in case you really have reference cycles, you can have multiple threads starting parsing from different points in that cycle, and getting deadlocked waiting for the DIE_IS_BEING_PARSED results from each other.
The only sane algorithm I can come up right now is to make the list of parsed dies local to each thread/parsing entity (e.g. via a "visited" list), and only update the global map once the parsing has completed (successfully or not). This can potentially duplicate some effort where one thread parses a type only to find out that it has already been parsed, but hopefully that is not going to be the common case. The alternative is some complicated resource cycle detection scheme.