I've always considered the use of sarcasm to be full of risks, especially online. There is an ethical pressure to not mislead people about what one believes. Yet, saying what you don't believe in an unbelievable way that's on the edge of undetectability is where a point can be made with humor. The moral quandary is that the risk is too high that someone believes what you wrote and they may rely upon your admission as substance to build their beliefs off of. This is especially true online if people reading what others write are non-native speakers and translating using a machine.

Imagine a machine-learning algorithm that was primarily trained on a dataset where people are often talking to themselves in ways where only they can detect what to invert (sarcastic). This AI will learn all kinds of nonsense and totally misrepresent the reality it believes the others are seeing and express.

That's the reason why I've started to add <Sarcasm> tag even though the hazard of someone not even reaching the bottom is high and never seeing it is high. :P