Blinded paths are a real improvement. They're not a magic shield. Don't let anyone tell you they make your node untraceable. Here are the actual attacks to keep in mind:

• Fee/CLTV correlation. An adversary watches channel policy updates across the network and tries to match them to parameters in your blinded paths. If you use your exact real channel fees in the blinded path, every fee update you make re-identifies you. Fix it: add a margin buffer above your real fees. The route blinding attack analysis covers this in detail.
• Graph analysis on small nodes. If you have 2 channels, there are exactly 2 candidates for your introduction node. An adversary can try both. Blinded paths help, but a poorly connected node is still a poorly connected node. More channels with diverse peers means better privacy.
• Payment probing. Someone can attempt to route payments toward the introduction node, fail deliberately at each step, and use the failure messages to narrow down where you actually sit. Dummy hops are the mitigation — they make it impossible to know how far along the blinded path the real recipient is.
• Introduction node visibility. The intro node KNOWS it's the intro node. It sees a payment arriving for a blinded route. It doesn't know WHO the recipient is, but it knows something blinded is downstream. Pick your peers with at least some trust. Don't use a peer you'd never trust with your routing business as your introduction node.