A lot of the early speculative articles do not appear to be accurate. The structural folks on an engineering forum that I'm on have it mostly sussed out (and there was plenty of speculation early on there, as well; some of the early theories have been disproven).
http://www.eng-tips.com/viewthread.cfm?qid=436924
The current running theory is that the means of transporting the bridge from its construction location to the final location placed an element known as "member 11" - the outermost diagonal on the side that failed - into tension, which is the opposite of what its eventual loading would be with the bridge sitting on its piers. Concrete does not handle tensile loads well. This was evidently known, and was compensated for by pretensioning member 11 (steel cables apply tension that allows the concrete itself to remain in compression), but the process of setting the bridge down on its piers placed that element back into compression. This meant the concrete now had to take up not only the compression from the dead load but also the compression from the pretension. This is believed to have led to the initial crack that was thought not to be an issue, and it wasn't, until the pretension was removed and the joint between members 11-12 and the deck failed in the process of relieving the pretension.
In addition to all of this, concrete is a poor choice to make a truss from. When doing the structural analysis of a truss, one assumes that the joints are "pin joints" i.e. the joints do not transmit any moment loads (torque / twisting). Steel can approximate that by bending slightly. Concrete is unforgiving.
No one knows why this wasn't built as a steel structure with a concrete non-structural deck, and decorative cladding around the steel members if someone so desired the bridge to look the way it did ... in other words, the same way as thousands of other bridges that are still standing. There was evidently a lot of politics behind the construction of this bridge.
No one also knows why traffic was allowed under the bridge during pretensioning (or relieving pretension, as the case may be). Pretensioning is recognised as being a dangerous operation, and in structures that rely on pretensioning to achieve their load capacity, the structure is not functional "as designed" until that pretensioning is complete.