How you define "your code is correct" might matter. If you perform any operation (not just floats) that may not be commutative in such a way that it could be influenced by something non-deterministic like thread execution, then I'd argue your code was never correct. (Unless of course your goal was non-determinism)
Typically, Ethernet network cards drop packets with invalid checksums, so you never see them. Of course, a checksum isn't a perfect filter, so errors can still get through.
Better yet, typically we use TCP/IP over our networks. TCP adds another layer of error prevention with it's own checksums, timeouts and retries: Error Control in TCP - GeeksforGeeks.
If I absolutely, positively had to know that the data had changed, I'd be always looking at a measurement timestamp. In this case, it's just a check to prevent a lengthy recomputation to display the results to an operator; if the equality check fails when it shouldn't we just lose a little time.