From my experience in the mammalian world (and this may apply to bacterial systems as well), it's not so much the number of genes in the plasmid as its actual size. The larger the construct is, the more difficult it will be to get it into your target cells in one piece, without degradation or shearing. Since the transformation efficiency is lower, you are getting fewer whole constructs per cell, so depending on how you've set up your promoters, the overall expression level can be significantly lower. The trouble with splitting your genes amongst two or more plasmids is that each will have different transformation efficiencies, and there will be a certain (perhaps large) number of cells that don't get the full complement of genes, interfering with phenotype analysis. And again, you'll also have to consider differential expression rates.
However, once you get your vector into the cells in one piece, theoretically they should all get expressed at approximately equal levels (assuming identical promoters). It's possible that steric hindrance among multiple transcription complexes may occur - I just don't know enough about bacterial transcription and the effects of circular plasmids to say.
One way to get around many of these factors would be to use a bacterial artificial chromosome or BAC. BACs are 7-10 kb vectors that can have inserts of up to 1 million bp cloned into them and are then electroporated into cells. One of (the many) cool things about them is they control their own duplication and partitioning at cell division, so succeeding generations should have essentially the same copy number as the original. They were heavily used during the Human Genome Project to amplify large sections of DNA for sequencing, but they are also used in many other studies, including synthetic biology. OpenWetWare has a list of some common ones, and NEB sells pBeloBAC11 systems.