Besides only doing training with the dog in as much of a drivey state as they're capable of and making everything as game like as possible, I also interrupt slow responses. I want and need silence to equal good for my dog's since they do compete in obedience so I can't just not reward the ones I don't like and reward the ones I do like or I'll have a very confused dog in competitions where I can't talk and there are no external rewards. What I do instead is to interrupt the dog the moment I don't like something with a cheerful
try again! or
too slow! or whatever blab comes out of my mouth as I reset the dog. Depending on what I'm doing I might also happily tag the dog and run away a few steps for the reattempt.
Example1: working on a fast down while heeling if my dog has 1 second to start moving down I'm not going to wait until the dog slowly starts to lie down at 3seconds and elbows hit the ground at 4. Instead at the 1.5second mark I'm already, happily, goosing the dog and moving away even if I can see the dog starting to bend. I personally feel that waiting until the dog already completes the behavior and then trying to give them feedback about speed or straightness or whatever is too confusing to the dog. Of course I start at a baseline of what I'm getting. If your dog's average speed is 3 seconds then I'm only going to wait 2.5sec-3sec to reset the dog...
Example2: crooked fronts. I don't wait until the dog is already sitting in front of me crooked and then tell them to straighten up. I feel that doing so encourages the dog to be shifty and adjusting position a lot if not given their reward right away. Instead I interrupt the second I see the dog starting to lower their butt for the sit but sidestep AWAY from the dog in order to interrupt the sit and at the same time make their error bigger. Of course in this example my dog should already know how to side step
