We used to do dog fooding in our company with our intranet product; after about 5 years our client wishes diverted so wildly from what we used it for that it began to work against us. We were actually building features for clients which interfered with the features we used/needed (and over 200 people in our company depended on) which resulted in a lot of time lost to make sure both everything worked well. In the end it was not sustainable; we almost went under because of it when we started work on the new major version.
I guess the answer is; it depends, but i'm very cautious now because of the above; once you have your whole organization and client base working with it fulltime, it is hard to switch and you often cannot say no to features as the competitors will be adding them. We should've (hindsight) just forked our intranet and forget about selling that part with the 'new' stuff.