The are several coverage tickets that run into a tcmalloc locking race condition. Most likely this is from the fork exec window here.

More details can found in SERVER-61860 which ran into the same issue (same stack trace) in the unittests.

It's theorized the normal the jstest shell doesn't hit this case because the lock window is pretty small already, but with clang coverage the code execution is slower because the extra overhead from generated coverage calls. More details in this comment.