If I understand the question correctly, you're thinking something along the lines of modular tertiary domains, but in a quaternary sense?
For a reversible interaction, heterotrimeric G proteins spring to mind. Once activated by a G protein coupled receptor (GPCR), the alpha subunit and the beta-gamma complex dissociate from the receptor and from each other. They can both go on to affect signaling by binding to various effector proteins, depending on the particular pathway in which they are involved, and their activities can be mediated by other proteins (e.g. the alpha subunit GTPase regulated by GAP proteins).
I spoke to my Protein Engineering professor, and she also directed me to a few examples, although I'm not sure they're quite what you're looking for in the strictest sense.
- Cyclin-dependent kinases, as @Armatus mentioned, bind to all sorts of different cyclins. (I get the sense that subunits found in multiple protein complexes will probably be more common in signal cascades. I don't have any real data on that, but it makes sense that a signal protein could interact with two or more protein complexes and even have slightly different functions in each.)
- Alcohol dehydrogenase (ADH) is an example of a protein with different quaternary structures between species. According to my professor, it occurs as a homodimer in mammals, but as a tetramer with some slightly different residues in budding yeast. I can't say whether the sequence changes are what affect the oligomerization, although my prof also mentioned "oligomerization domains" that could mediate protein-protein interactions.
- She also directed me toward a slightly more interesting example (in my opinion), though, again, it's also more along the lines of oligomerization. Small heat shock proteins (sHSPs) undergo something called "dynamic subunit exchange" and can take on a wide variety of oligomeric architectures. Binding is transient, but it's still kind of neat. There's a review here: http://www.sciencedirect.com/science/article/pii/S0968000411001903. Figure 3 shows different oligomeric architectures from the one monomer.
It seems that pretty much any protein that interacts with two or more separate protein complexes could be considered an answer. Neither I nor my professor have managed to come up with an example of an irreversible assembly of common subunits, but she is of the opinion that, in all the different possible protein structures, there is probably at least one example of everything, no matter how outlandish it seems.