You are right with the distinction, I had data race conditions in mind.

Race conditions can well happen in a correct C/C++ multi-threaded program in the sense that the order of specific computation steps is sometimes random. And for operations such as floating-point addition, where order of operations does matter, the exact result can be random as a consequence. But the end result should not depend dramatically on it (which is what the poster at lockdownskeptics.org claims).