Why illumina have staggered the release of human methylation arrays is unknown, most likely an in house decision. I can speculate that it may come from a number of reasons, most likely not to do with the number of probes per say, but likely the content of the probes. The bead array technology hasn't changed across all 3 releases of the human methylation series, so it's not to do with "how many beads can we fit on this silica chip", but more likely on characterising, testing, and validating all the capture sequences and where they target. This will more than likely be the last release we see of the humanMethylation chip series, as sequencing will become a cheaper and more profitable process, in a similar fashion to what's been happening with gene level expression.
Adding more probes adds to the price. If you add more probes, the array becomes more expensive. Thus, it only makes sense to add more probes if the additional probes are informative. It's a lot of work to find millions of informative probes. Also, the arrays are mostly designed for clinical applications and they are sufficient for that purpose.
If you want all the CpGs in the genome, then you can do WGBS, so there is no need for an array-based solution.