There's no data anywhere to prove that soldered RAM has a smaller failure rate than RAM that is not soldered. Their might be some data somewhere, but whoever has it has certainly not shared it with the world.
I would be willing to bet that the number of instances where a RAM slot connector has been involved in RAM component failure is not statistically significant enough to save Apple enough money to list as a separate line item in their Annual Report to shareholders.
The main, and most likely, only reasons the RAM is soldered on now has a lot more to do with thinness and lower latency than it has on any other potential fringe "suport cost" benefits. ... and the thinness reason is kinda stretching it, TBH. With their supposed design prowess, they could have found a way to mount RAM slots on the edge of the logic board, side-by-side without compromising thinness. I suspect it was the lower latency that was the driving force behind moving to soldered RAM.
'Forcing" (I really hate to use that word here, btw, as everybody still has a choice) customers to purchase upgrades at time of initial purchase is another possible fringe "benefit" for Apple, but like the move to soldered RAM, there's no data to support that conclusion out in the public ... just assumptions.