I read an account several years ago about a K-9 training program in the US Army during WW2, that I don't see mentioned in the linked site. As it never produced a successful deployment, perhaps it doesn't qualify. Or perhaps the site author(s) are simply not aware of this rather peculiar effort.
The program was based on the assumption, not well grounded in any science, that dogs would be able to distinguish Japanese from other people based on the smell of their race.
A unit of nissei (first generation American-born, with Japanese parents) volunteers were recruited from the west coast displacement camps for "special" service. They were placed in a remote camp in the Southeast (I seem to recall it was in Florida, but I could be wrong on that).
The concept was to train dogs to seek out racially Japanese individuals in overgrown tropical wilderness areas, and to rip their throats out.
A group of white soldiers were the dog handlers. The dogs went through a sequence of training that included feeding on raw meet placed on the necks of the Japanese volunteers (no sh!t), then tracking and finding (and distinguishing) the Japanese volunteers in the wilderness, and then learning to hate the Japanese volunteers. This was the sequence of the training.
While the dogs showed a relatively good ability to track and find individuals they knew (and liked) in the wilderness, they showed relatively poor ability to distinguish Japanese from non-Japanese among individuals they did not know in the wilderness.
The nissei volunteers seems rather dubious on the whole affair from the start -- feeding the dogs off of their necks generated quite a bit of discussion and concern, but most were patriotic volunteers who were committed to doing their duty for the war effort, and the dogs were nice (and careful) enough. You actually get kind of bonded with a dog when you let it eat just inches from your face. The games of hide-and-seek with the dogs were all good fun for 19 year olds away from home for the first time. But the deliberate abuse of the dogs in the final stage of training, to get the dogs to hate anyone who was Japanese, caused a lot of objections among the volunteers.
In the end the whole program was shut down after about a year, as it never was able to demonstrate that dogs could be turned loose and counted on to reliably distinguish between white and Japanese scent trails in tropical brush, or anywhere else for that matter.
Might have been possible if they had tried to train the dogs based on scent characteristics of actual enemy soldiers in the combat zone(s). My understanding is that this has become the standard tactic for K-9 combat units. Whether it is differentiated components of their diets, their uniforms, their gun oil or even their personal hygiene habits and products, soldiers from different nations generate unique scents that dogs can learn.
But the concept that they would have a unique scent based just on their race seems to have been a poor premise for a training program.
-Mark
(aka: Mk 1)