The reason for that is that YouTube has added an A/B/C testing feature for creators to try out multiple thumbnails at once. It tracks which of them performs the best over a period of time, and then implements that one after the testing period is over.
Overall, my feelings on the feature are mixed, but I do love new little toys to play with, so I've been messing around with it and probably will keep messing around with it for the next few months. Above, you can see the three thumbnails we're going to test for the Windhelm unemployment video tomorrow. My gut says that either #1 or #2 will do best, but I wanted to throw in a "bad" one as well just to see if it really does make a difference. Which do you think will ultimately be most effective?
Longer term, I think I'll go back to trusting my gut and making thumbnails that feel good first and foremost, but this tool is definitely a good way to sharpen the skills and check in with the audience. Best not to worry about it too much though-- time and energy is usually better spent just making another great new video, rather than obsessing over every little color and shape in a thumbnail.
Austin