Do you design it with a DoS prevention in mind?
That's definitely a must have, for a system like this - there will be a big money trying to break it.
Recently I was reading a bit about how darkcoin solved DoS problem for coin joining; they do it by taking a collateral from the coinjoin parties, so if any of them would back off he gets to pay for it. That's a smart DoS prevention idea, though I'm guessing people also came out with other solutions.
But going back to secured chats - they also need some DoS prevention, especially if the client code is in javascript.
We do think about dos protection and other attacks. For now we are doing a reference implementation that will be hardened for production, so we are trying to have something simple where parts can be replaced, for example the channel system could be replaced by something like bitmessage, irc or in some cases the bitcoin network itself (systems offering more guarantees). We will rather adopt systems with full teams dedicated to those problems than reinvent the wheel.
On the same line coinjoin protocols can become more ellaborate and offer greater protection, anyways we will try to keep the modularity and flexibility. For now we rather have things running and people using it than waiting for the perfect system in face of threats that may not materiallize or people can avert in other ways (now people can switch channels for example).