According to what Martin discovered from the wxGurus, these won't help us, though... as it's wxString that is unsafe.
Apart from that, I've read a lot about and implemented several lock-free algorithms, and while they're on several orders of magnitude faster in the right conditions, they likewise perform miserably in the wrong conditions. Also, they are by no means easier than locking schemes, or more resistant to failure. Contrarily, they bring a whole different set of problems, and errors can be hard to anticipate or detect.
In fact, lock-free algorithms are not even sufficiently understood by many who think that they can write articles about them, as you can see for example at Dr. Dobb's Journal.