In general, hardware products that appeal only to hackers (OpenMoko, Neuros, OGP, etc.) fail while hackable mass-market hardware (G1, WRT54G) succeeds. You can't ship a "rough" product and expect the community to polish it, since polish is what the community is worst at and peer production is so slow that your product is obsolete before the community can finish the firmware. Also, hacker hardware is expensive because it lacks economy of scale.
Their hardware was epically bad. It used an ancient and buggy GPRS radio (1.5G, ~5KB/s), had naive drivers (no power management!), the sound chip didn't really work, the shitty resistive touchscreen was recessed so you always needed a stylus... It was not feasible to use it to place or receive telephone calls.
The software was just as bad, and was completely rewritten every couple months. It was based on each of [GTK+, QT, Enlightenment, DirectFB] at least once apiece.
I don't see how they could have possibly done a worse job.
It seems like they would have been well positioned for exactly this sort of thing. I'm not sure if they had problems with the hardware or whether they just couldn't get traction.