You can price this relatively quickly, and see that it's probably not worth it.
~1 week outage once every ~5 years, let's say you lose 100% of revenue that week, that's 0.38% of revenue. Now most likely they're not losing 100% of revenue, demand for cars is not very elastic... it starts to get pointless all around.
Which is frustrating for consumers, but the profit for perfection isn't usually there.
I think you'd have to compare that .38% of revenue to the cost of developing the paper process, as well comparing what revenue you'd lose with the paper process anyway. It doesn't seem crazy to me, unless your paper process is hard to develop.
There's also an argument that developing a backup process is a great exercise to help people understand all the parts of your system and how they interact, as a first step to making them more efficient/redundant/secure/delightful.
~1 week outage once every ~5 years, let's say you lose 100% of revenue that week, that's 0.38% of revenue. Now most likely they're not losing 100% of revenue, demand for cars is not very elastic... it starts to get pointless all around.
Which is frustrating for consumers, but the profit for perfection isn't usually there.