Assuming that the systems are truly encrypted end-to-end, how do we know for sure that noone can actually decrypt it? I mean, apart from the Eng team that directly works on that part, are we basically trusting them at face value that it works? Or does this code need to live in an open-sourced environment so as to make it more trustworthy?

To be clear, I don't mean trusting the encryption algorithm, that part is easy because math. I mean trusting that that algorithm is actually what's being used under the hood and actually noone can decrypt it for any reason.