I could reproduce the problem on Windows using the 1.000.000 virtual threads test case -- a hs_err_...pid file was also produced. Would it help if I share it with you?