Meh. The logic still applies: on a bad network, the first request can hang for half a minute when a retry would get lucky and go through immediately.

Don't handcuff your users for no benefit, just because you can't write a well-behaved backend.