We found this dog wandering through the neighborhood. We fed the dog, our neighbor found the owner and returned it, so this story had a happy ending. Unfortunately, I've seen some "dog" projects that didn't have happy endings.
These dogs managed to escape the fence (aka the governance process) and ended up on my desk. The sales team thought these "dogs" were cute because they required minimal sales effort. I recall one project in which the customer's entire systems support team was leaving so the conversion needed to be finished in six weeks instead of the usual six months. Many of my "dog" projects entailed functional requirements well outside of the TPA's technical and administrative capabilities, so extensive system modifications were required. Of course, the modifications could not be completed in time for the conversion. No problem - just implement a few workarounds!
If you want to keep the dogs off your desk, you will need to patch your governance "fence." Design a point weighted scoring worksheet so the sales team can quickly profile a project to determine its complexity and match for current capabilities. Develop solid cost estimates and stick by them when challenged. I recall spending four months defending a project cost estimate because it was 50% higher than the sales team estimate, and it turns out the final cost was much higher than my "excessive" estimate. Finally, if you have no choice but to take on a "dog" project, make sure the project sponsor understands the risks and is prepared to secure the proper resources to complete the project.