Shadowbans are a very useful tool against human-powered disruptive activity (sockpuppetry and what-not), because they shift the effort asymmetry to favour the attacker a bit less than it tends to do otherwise. (Given the possibility of being shadowbanned, the attacker has to either constantly check that the fruits of their labour are still accessible, or risk wasting work posting into the void on an account that has already been null-routed)

That this technology, when available, invariably winds up being used against good-faith users seems to be the usual grubby-fingers problem where the powerful take away from the commons of social trust for a temporary advantage, akin to police using COVID contact tracing data to chase small-time drug dealers.