Why does TTG allow suppressing hiss and the sound of wind hitting the microphone, but prohibits cutting out a cough or a dog’s bark or a baby’s cry?

Because the determinant is “what was heard at the scene.”

• Hiss and the sound of wind hitting the microphone are not heard at the scene by someone standing next to the camera; they are entirely due to the microphone/camera setup (in other words, if there were no camera at the scene, camera-generated hiss and the sound of wind hitting the microphone wouldn’t exist).

• But a cough or a dog’s bark or a baby’s cry would be heard at the scene by someone standing next to the camera (in other words, those sounds would exist even if there were no camera at the scene).

The exact same principle applies to video as to audio:

TTG allows the correction of lens/camera artifacts like lens flare, internal reflections, digital noise, and curved lines caused by barrel/pinchushion distortion, which are not things that are seen by someone at the scene—

—but V4 prohibits any changes (other than color correction) to anything that would have been seen by someone at the scene.