I'm not sure that's true. It would certainly be an interesting anthropological undertaking to pin down the details, but (at substantial risk of being wrong) I feel like this was an assumption in early attempts at abstraction and we only feel the need to specify "non-leaky" because we have discovered important things that previous attempts have - in practice - tended to leak.
I think I'm in agreement with you here—we're slowly, as a community, discovering how to make non-leaky abstractions. They've been around for a while in places where "in practice" had a lot of legroom (pure mathematics). CS has a lot of legroom too, but it's taken a long time for us to think about it in such a way to know where we can place our weight.
We may get there one day, but for right now just spec'ing interfaces as combinations of consistent laws is a stretch.