There are two well known example related to the regulation of galactose metabolism in S.cerevisiae and C.albicans. These cases are well described in the article Martchenko et al, 'Transcriptional rewiring of fungal galactose-metabolism circuitry'. Curr Biol. 2007, and also in Rokas et al Transcriptional Rewiring: The Proof Is in the Eating, Curr. Biol 2007. There is also some description in chapter 3 of the book "The Origins of Evolutionary Innovations" by Andreas Wagner, and in these slides I did for a journal club on the same book.

In S.cerevisiae, GAL4 is involved in the galactose metabolism, regulating the expression of other genes that are able to convert this sugar. In C.albicans, the regulation of the same enzymes is done by other genes (one of which is CPH1), while the homologue of GAL4 is associated to the telomer, and is involved in an unknown function. At the same time, the homologue of the gene CPH1 which regulates galactose metabolism in C.albicans, is involved in mating in S.cerevisiae.

So:

1. GAL4 -> involved in galactose metabolism in S.cerevisiae, involved in an unknown function related to telomeres in C.albicans.
2. CPH1 -> involved in galactose metabolism in C.albicans, but involved in mating in S.cerevisiae (where it is known as STE12).

image taken from http://www.sciencedirect.com/science/article/pii/S096098220701562X#gr1

In general, I think that you are looking for homologues that have different function, your best bet is to search among transcription factors. A small mutation in certain positions of the sequence of a transcript factor can have a huge impact on the specificity of the transcription factor, making it recognize a completely different sequence. This leads to a change of function of the transcript factor, while the rest of the sequence remains conserved.

Identifying a retention of part of a common function might be tricky, as "part of common function" isn't something easily captured by Gene Ontology. If you need some examples, there's a plenty of literature on the topic (Iike in particular the work of Marc Robinson-Rechavi, so here's one of his recent papers: http://bib.oxfordjournals.org/content/early/2011/06/16/bib.bbr031.full ). However if you need some large scale data, I would start with "clans" of protein domains in Pfam. Basically they contain related, but functionally diverged protein domains. If you map that onto orthology classification (let's say Kegg Orthology), this would give you several good starting points to start a detailed search from.

Rather than Pfam Clans, I suggest you might use the Panther resource. It covers all of the reference genomes as defined by Quest for Orthologs. Not only does it use phylogenetic trees when defining families, it also applies GO terms to them, so I think this is a really good fit.

In their 2011 paper "Phylogenetic-based propagation of functional annotations within the Gene Ontology consortium" there is a very nice example of exactly what you describe, where a family of super oxide dismutases has lost function (to become chaperone proteins) in one of the clades (PTHR10003) - see Figure 5. The paper also describes how different automatic GO annotation methodologies (InterPro2GO, PAINT and Compara2GO) assign terms.

Hope this helps! (Thanks to my team for providing examples :-) )

If protein interactions work, then BioGRID is a nice place to look at. It gives all the list of protein interactions which have been conserved during the course of evolution. And I would trust physical interactions and not functional associations, as the former have been in wet-lab by techniques like Yeast-2-hybrid (then again, they do contain false positives).

You can either download(link) all the interactions from this website or just search for the proteins you need. As an example I searched for Hex protein, and the results can be found here in this link.

How about Globins in Archaea and in Human ? Enough to be classified as "orthologs with functional divergence" ?

Try treeFam database. Best!