If there is indeed a degree of uncertainty why do nearly all the models diverge to the warm side?

Models should use input  to generate a range of scenarios around the middle of the range.

At the moment we are in the edge of a fading La Nina but it is still cold. Hence the starting point is this edge.

Should we expect that a La Nina can only develop after an El Nino?

If not then the predictions are wrong unless they include a spread that includes a reasonable chance of La Nina, more than 25% as this would be the historical likelihood.

It is not appropriate, ever, to say that a fourth La Nina cannot occur.

La Nina is more like a four side dice than a two faced coin but ruling it out due to previous events is wrong.