True, but OP's threat model doesn't involve state actors outside Indonesia, so traffic analysis of the "last mile" between Mullvad node and whatever non-Indonesian service OP is trying to use (Twitter, Discord, …) is not really relevant here. (Assuming Indonesia doesn't have capabilities we don't know of.)

What might be more interesting is the case where the Indonesian government forces Twitter/Discord to give up IP addresses (which I find hard to believe but it's certainly not impossible). But then they'd still have to overcome Mullvad. It's much more likely that if OP has an account on Twitter/Discord, it is already tied to their person in many ways, and this would probably be the main risk here.