NEAR Intents added additional 9 to the reliability of the protocol (now it's 99.99%). Now, we're 10 times more reliable comparing to the closest competitor to our best knowledge.

Preceding the question: for multichain protocols fails are normal, since different chains are async in their nature. You can start the swap from one chain and then... Other chain is down. Or their RPCs are down. So there should be a way to push the execution forward. Even if the protocol itself is decentralised! Decentralisation and reliability are different dimensions.