Every other payment method you consider not to be a failure today is the result of the same series of pre-release testing, years of trial and error, and still - not perfect. Lightning works, when it does, as has been documented by many. Even the detractors have to admit that it works on certain thresholds. If you need something more certain, stick to on-chain Bitcoin. I'm actually pretty sure you'd still prefer to attempt Lightning again, if offered to fly and travel by camel to pay whoever it is you want to meet to pay.
It's like schrodinger's cat - create a channel, find out it's useless, close channel. Day later, open new channel, find out it's useless, close channel...day later... (I'm defining useless as every payment failing due to node errors of various kinds).
Schroedinger's cat does not work like that. There's no paradox in LN.
The paradox is that (as far as this user can tell) it is not possible to know the properties of a node (expiry length/transaction fee) and set your wallet to meet those properties. This gives rise to 'hit and hope', or 'is the transactional cat in state a (dead) or state b (alive)'?
I can see the potential of it, and plenty of the elements do function as expected, just the final actual transaction bit seems 'slightly' error prone.
To put this in context, I'm giving up on attempting a $50 transaction after creating three channels and initiating six merchant transactions (x2 per channel), each of which were rejected themselves multiple times.
Edit: Now up to 6 channels. Two days and counting now, still unsuccessful.