In your description I still take away that the root problem isn't in the technology, but in the engineer's implementation. Even a calculation that is guaranteed to produce the correct results [if implemented poorly or with errors or with exclusions] can fail.

The engineer who wrote the contract appears to have made a critical logical error.

That does bring me to your conclusion, though. I've paid less attention to the marketing and promises and have assessed the technology as-is (within the confines of my knowledge) -- as such I see what's possible instead of seeing it as a product that under-delivers.

I wonder if a lot of people (especially on HN) would view it differently if it weren't for such bold claims on the developers' part. I mean, I think you're probably right in it being naive in its guarantee of predictability. I think any actors on the platform should assume its supposedly perfectly predictable behaviour is not quite so when they take to designing software or a contract. But that's part of what I meant when I've said it comes down to engineering practice -- when it comes to critical systems assumptions just shouldn't be made in young technology no matter what anybody tries to sell you.