1) Xnodes are fully decentralised
Any wallet can run as an Xnode, not just a few semi-centralised clients
Xnode setup is not complicated or time-consuming
2) Xnodes are a trustless design
An earlier iteration of XC's design specified a dynamic trust system; this has been improved upon with a fully trustless design
Rev 2 Xnodes will use trustless multi-path multisig
3) Xnodes work
Early tests of Xnodes' mixing function (Rev. 1) have not resulted in any kind of flaw in this aspect of XC's anonymity solution.
Our not-so-well-intentioned friend Chaeplin claimed to have found a "design flaw" in Rev 1. However he fails to apprehend (a) precisely what was being tested, (b) what would constitute XC failing the test, and (c) what his results actually show. For a summary, see
https://www.google.com/url?q=https://bitcointalk.org/index.php%3Ftopic%3D630547.msg7310485Tests of Xnodes' initial multipath implementation (Rev. 1.5) have turned up no problems. In fact, no one was even able to mount a claim for the bounty.
Multi-path is being fully implemented as we speak for Rev. 2. When it is released, there will be (a) a sizeable public bounty, and (b) consultations with several cryptographers and other experts.
Given past progress, we're pretty optimistic that Rev. 2 will be delivered on time, and will compound the anonymity that is already working in XC.
4) XC is already anonymous, and working
I can't stress this enough. It's working! It doesn't mysteriously fork. It was released on time. There have been no setbacks.
XC is only 5 weeks old, and we're already at this level.
XC is not a completed design. It anonymity will compound. Its user-friendliness will become a 1-click thing. Thorough, professional testing will render it bulletproof. See the roadmap:
https://bitcointalk.org/index.php?topic=630547.msg7535822#msg7535822.
Xnodes do not steal coins. But for the sake of historical accuracy, here's where this rumour started: Rev. 1 was released to test whether anyone could prove that a link exists on the blockchain between sender and receiver. Chaeplin misconstrued the purpose of the release and harped on about the fact that because mixers forward coins, coins could be stolen. This is true, but it's besides the point. All we were testing was whether anyone could provide a link on the blockchain. The rest of the build, including vulnerable transaction-forwarding, was just scaffolding so that the tests could be carried out. But, alas, Chaeplin appeared not to be amenable to realising this fact. Perhaps if his intention was to contribute constructively he would have avoided this lapse of comprehension.