First failure seen here (v6.10 nightlies), can reproduce locally on master.
valgrind-helgrind does not help due to the very high number of false positives (mainly due to tbb). valgrind-memcheck does not report anything helpful – a few leaks here and there, mostly in TObject, that do not seem related to the issue and that ROOT always gets.
clang -fsanitize=thread is confused by our TSpinLock and reports several false positives, did not see anything useful there either.
I was not able to reproduce the crash inside gdb yet.