Is the necessary that stamens have two microsporangia per theca? No. Many species in Cryptocarya group of Lauraceae have microsporangia that have fused such that they have only two microsporangia instead of the typical four (Rohwer et al. 2014). I interpret their description as saying that the microsporangia within a theca but it's not clearly stated as such.
Within the Magnoliidae, one pollen sac of a theca is lost is some taxa (Endress et al. 2008). Evolutionarily, the Hamamelidaceae have reduced the number of pollen sacs from two to one (Endress 1989). I could not find examples of plants with more than two microsporangia per theca.
Why do most plants have two pollen sacs per theca? It's probably a result of the evolutionary history of plants in general because it is so common. Even those that have only a single pollen sac seems to start out developmentally with two. It's probably due to the underlying genetics.
The genetics of floral development is complex but being slowly teased apart. The overall development of floral patterns is set by a group of developmental or regulatory$^1$ genes called MADS genes. Collectively, the MADS genes (along with other regulatory genes) control the overall organization of the flower structure. For example, MADS genes determine the arrangement of floral whorls from sepals (outermost whorl) to petals to stamens to carpals (innermost whorl). Changes in use or timing of the MADS genes can contribute to different types of floral morphology, such as the loss of sepals or big, showy sepals but no petals.
Some of these same MADS genes contribute to the formation of the stamen structure. The review by Goldberg et al. (1993) gives a thorough but now dated review of the developmental genes that control anther development. The figure below (their figure 3) gives a nice overview of the stages that lead to development of the theca with two microsporangia and ultimately to the final stage of fertilization. Even more specific, Bowman et al. (1991) show that AGAMOUS is one regulatory gene that is active during the development of the theca and microsporangia.
I can't speak in detail about how regulatory genes work in flowers (I know a bit more about regulatory genes in animal development, especially from an evolutionary developmental perspective). In animals, regulatory genes often work by concentration gradients, both by the concentration of the regulatory gene itself, by the concentration of other regulatory genes, and by the concentration of the regulated genes. This process helps to determine the anterior-posterior axis in animals with bilateral symmetry, as well as the left and right sides.
I suspect that floral regulatory genes work in a similar fashion. Changes in regulatory expression can lead to bilaterally symmetric flowers (think of mint flowers, monkey flowers, etc. See Hileman and Cubas (2009 for a good review.). I suspect it is some type of symmetric expression of regulatory genes that leads to the formation of two microsporangia per theca. Mutations that cause changes in the expression patterns of other regulatory genes can cause the initial expression but subsequent loss of one of the microsporangia, as noted for the Lauraceae and the Hamamelidaceae.
Broadly but simply, regulatory genes are genes such as transcription factors that regulate or control the expression of other genes.
Bowman, J.L., et al. 1991. Expression of the Arabidopsis floral homeotic gene AGAMOUS is restricted to specific cell types late in flower development. The Plant Cell 3: 749-758.
Endress, P.K. 1989. Aspects of evolutionary differentiation of the Hamamelidaceae and the Lower Hamamelididae. Plant Systematics and Evolution 162: 193-211.
Endress, P.K. and L.D. Hufford. 2008. The diversity of stamen structures and dehiscence patterns among Magnoliidae. Botanical Journal fo the Linnean Society 100: 45-85.
Goldberg, R.B., et al. 1993. Anther development: Basic principles and practical applications. The Plant Cell 5: 1217-1229.
Hileman, L.C. and P. Cubas. 2009. An expanded evolutionary role for flower symmetry genes. Journal of Biology 8: 90.
Rohwer, J.G., et al. 2014. A phylogenetic analysis of the Cryptocarya group (Lauraceae), and relationships of Dahlgrenodendron, Sinopora, Triadodaphne, and Yasunia. Phytotaxa 158: 111-132.