I checked this with Ivona (the new voice that we're A/B testing right now) and you're right, she does say it differently as a separate word. It's because the grammar rules for the TTS are wrong. There is no word vakt that is pronounced like that, but there could have been if the word had worked like some other words. The pronunciation with en is the right one.
It is, yes. Old English went through a sound change that made some hard k sounds into tch sounds. Another example is "leech" (an older word for "heal") and läka in modern Swedish.
In some cases, modern English actually retains both versions, often because either fell out and in of fashion or was regained from other languages. With "watch", we also have "awake" - they're both from the same Old English root, but have developed separately.