The call to ResolvedMethodTable::unlink() grabs the ResolvedMethodTable_lock during parallel cleaning.

However at the same time this lock may already been held by e.g. compiler threads trying to resolve the method. (e.g. ResolvedMethodTable::add_method).

Backing this out due to other concerns about the changes which might need more discussion: timing uses another time base (us) with another data type (jint) compared to all other collectors.