The reason why it works is text messages are sent on the same control channel that your phones use to "negotiate" air time for voice calls.

The average cell phone control channel runs at 9600 baud (or higher). A 120 character message carries as much information as a 30 second voice contact. I'm not talking bytes, I'm talking about logical information. So, assuming their are 100 phones trying to get a voice channel, and there are 128 time slots available on the data channel, and no voice channels available. Those 100 phones are going to get a DENIED message, while your message, which isn't asking for a voice channel, will have its message processed and routed. Oh, what about if all 128 time slots are currently filled at this point in time? Your phone will wait and try again.

Cellular companies would never do this, but there are a few technical reasons (not software, hardware...) it be difficult, but if the cell phones could "revert" to simply two-way simplex radio devices, that could be useful. Instead of a star-topology wireless network system, it would become a distributed network system. Amateur Radio is valued because of its distributed nature (yes, I know we have repeaters, but you can still effectively operate with most mobiles, without a repeater). Kazaa and Napster were/are valued because of their distributed nature.