Someone had a theory that the AI is now in charge of the updates and decides on its own who will enjoy a nice run of traffic and who won't every weekend. Interesting theory if you ask me.
I posted this but I'm pretty sure others have as well before.
It's pretty much because this is how ML/AI actually works.
I was involved in a few (non-search) projects that involved machine learning and a crucial part of ML is to let it run its course and not manually interfere with it too frequently.
Basically you tell it you want it to generate these results. The ML algo will then attempt to replicate those results by adjusting different factors on its own and evaluating the results of those changes and then implementing new changes until it gets to a point it generates results that are close to your initial specifications.
But the above involves you letting it do its thing on its own for a while - something that will inevitably mean mistakes and bad results in-between - until you get the result you want. If you interfere too much and too frequently manually then the algo won't really "learn" on its own, so to say. With every manual interference you kind of mess with the learning curve. This is very oversimplified but it gets the point across.
Core updates could be when they do significant manual changes to the algos. It's why they are probably done so infrequently. They have to let it run some time before they can interfere.
The seemingly weird search results and very frequent "updates" in-between could be just the ML algos working and constantly adjusting ranking factors to try and generate the desired results.
