Wasn't joining all worker threads before continuing manager thread, was fine unless trace load was cancelled midway, since some threads might continue for a while one then. Added explicit join before existing manager thread then, for all worker threads, and all instability resolved.
↧