genomics - What are the limiting factors for gene length and number of exons?

This question drops firmly into the lap of molecular evolution and the constraints that are placed upon genes by the forces of mutation, selection, drift and recombination.

There are numerous situations, particularly gene duplication, that can result in a gene that is free from the selective constraints of it's parent, many of which will accumulate so many deleterious mutations as a result of stochastic processes that they will become non-functional e.g. psuedogenes. Some can be altered and rearranged, accumulating exons and introns, and if they infer a fitness benefit on the organism, may be moved to fixation within a population.

Evolution is a population genetics process, and there are many variables which can effect the outcome, not least the difference in populations size. The genomes of larger populations (such as those of bacteria) appear to have much smaller genomes, and of course no (at least not spliceosomal) introns, perhaps as a result of increased fitness due to the decreased generation time of an organism with a more slender genome. It would be a good idea to read The Origins of Genome Architecture by Michael Lynch, as I think he answers your questions, better than I can.

Many of the genes you retrieve from EnsEMBL will of course have experimental evidence to support them. The genes that are predicted in the pipeline can be looked upon with less confidence, but you can of course look at the alignments with closely related species to see if you think the introns/exons are indeed viable. An example of a gene with 79 exons is the Dystrophin (DMD) gene, the longest annotated gene at 2,217,347bp (see Roberts et al, 1993 and Nishio et al, 1994).