I mean, even in low-latency trading where error checking is viewed as a sign of weakness (I kid, mostly) at least people have the common sense to add sensible l

I mean, even in low-latency trading where error checking is viewed as a sign of weakness (I kid, mostly) at least people have the common sense to add sensible limits to message size (like MTU) when using length-prefixed protocols. Even in these situations where the I/O is across all-ASIC optical and electrical paths corrupt messages do happen.