That's a valid point. The friction is relocated, not eliminated. The channel liquidity problem is real and anyone running a node knows it.
What I was getting at is the end-user psychological shift, the person sending sats doesn't perform the calculus anymore. Whether that calculus moved upstream to routing nodes or LSPs is a separate question.
But you're right that calling it frictionless is imprecise. Friction-abstracted is probably the better term.
