It's not intended to be ultra resilient to - for example - the NSA, it's designed that way to ensure we (as the operators of the backend infrastructure) and the client we built if for (a government department) cannot ever accidentally collect or expose any PII related to the platform users.
Without pushing changed client code (mobile apps, in this case) or MITMing you proactively before the TOFU exchange, even though I have god mode access to all the infra, I cannot (and can prove I cannot) reveal anything more than IP addresses and guids/times/sizes of user to user comms.
If my database gets popped, I and my client have no data breach reporting obligations. And I sleep _much_ better at night knowing that.
(If LOE show up with a warrant, we'd happily hand over the IP address/connections times/guids/guid-social-graphs - and they'd be able to go elsewhere to try mapping IP addresses to identities. So I'd strongly advise against running your drug business or terrorist cells on the back of this project. But _I_ cannot tie identities to guids. Unfortunately, I and this project are also from Australia, so if the App Store/googleplay binaries are already backdoored, I could soon potentially get put in jail for saying so...)