Direct gas collapse would work if there was little angular momentum in the region compared to the overdensity that starts to collapse. I'm sure this has been simulated, how probable is that?