Today, when I again woke up unreasonably early I realized I have been doing it rather constantly every week. To make sure it wasn't just some confirmation bias I needed some more data to back this theory up.
I don't sleep with my watch on my wrist. I just feel it's distracting to have anything attached to myself while I sleep. So instead of monitoring my sleep quality I focus on improving the quality of my sleep. But even if I don't wear it at night doesn't mean I don't have any data of my sleep patterns. Obviously my bed time shows up on daily activity charts as "not worn". So I at least know the times when I have gone to bed and got out of the bed. Given that I usually fall asleep quite fast (after reading about half an hour) and get out of the bed usually as soon as I'm awake I know pretty accurately how long I usually sleep as well.
Quick check on that data did indeed prove my initial observation. During the past three months I checked most of the weeks it has been Thursday morning when I have woken up the earliest. The few other weeks the earliest morning was either Wednesday or Friday. So still around the same time of the week. Weird.