Though most of us have been taught that friction is independent of the area of contact, according to Bowden & Tabor [The Friction and Lubrication of Solids (Clarendon, Oxford, 1950)], there is an area dependence, but on the very small scale. This is because on a microscopic scale, there are no true, continuous surfaces, so two surfaces pressed together are actually touching at a very large number of microscopic points, at different pressures, to boot.
Nature had a good
article (sorry - subscription required) which made the analogy to Velcro, where the number of points of attachment (velcro hooks) is proportional to the applied pressure, and where some points come apart sooner than others (since they don't all see the same load).
Can this be extrapolated (scaled up) to our question about wheels vs. treads to show that for very uneven surfaces, compliant treads offer more points of contact, and less likelihood of losing contact altogether? Or, as Matt B. said in the linked thread, "because the wheel treads are actually 'hooking' the carpet."