I'm not convinced of this. At best, shadow-banning wastes a little bit of their time and delays slightly their inevitable creation of a new account.
> If you instantly ban them, its easy to spot, easy to automate and doesn't really solve the problem.
Shadowbanning only really works for unsophisticated users. If you know it's a possibility, it's easy to check for (just create another account and check with it). Phisher's are probably going to be fairly sophisticated, since they're working to make a profit, not to only disrupt or push a POV.
I run a service that spammers and pfishers love. I started by banning these people (and sending them an e-mail letting them know why). They didn't stop, they just grabbed another free e-mail and continued. Now I quietly just delete the data they submit and never ban accounts. I let them submit data, tell them its accepted, then some time later (1-120 seconds) just delete the data. Their account still works, they just can't do much with it. Once the system became unreliable for them, they just moved on to something else.
I'm under the impression that spamming isn't as easy as it once was, and a fair amount of effort needs to be devoted to bypassing spam/abuse detection and blocks, which is exactly what a shadowban is.
> Once the system became unreliable for them, they just moved on to something else.
I would imagine that would depend on if there are equivalent services that the spammer could more easily migrate to. It sounds like they didn't move on from spamming, just from your service. Switching to drop in replacement would definitely be easier than writing a shadowban detector.
I'm just disputing that shadowbanning is a more effective solution than straight bans. It's just not that much harder to detect.
If that's too slippery slope, then so is the email _DMARC override settings.
And doesn't this kind of behavior open them up to a lot of legal risk? By moderating _DMARC records, they're assuming responsibility for email content going into the future. If inappropriate content is then passed via email, and Cloudflare failed to blackhole it with _DMARC recordsy, then the can't they become liable for that content?