They certainly are counting views on the backend also, and I'm sure they know exactly what the cause of the discrepancy (or "drop" as they term it) is.

They probably use a combination of the API and raw server requests due to how easy it would be otherwise to spoof viewership for ad revenue fraud. Would not surprise me anyway.

I also see the opposite problem: can one abuse that API to artificial inflate the view count?

It does kinda make sense for once, you probably wouldn't want to just count API calls for views. I heard you need to watch a significant portion of the video before it counts as a view.

I realized this when I watched one of my friends music videos to give the extra view (they had less than 100) but the views number didn’t go up because of my ad blocker.

afairc sub-100 views are not counted in realtime anyway.

When I watched it without the ad blocker the views went up by one. I feel like it's a lot more realtime now because I swear I see the view counter updating as I'm watching some videos.

I think it was 300.

Then it used to stop at 301 views, while the system was verifying the veracity of the views, or at least, that's what I've been told…

Ad-block views don't help anyone anyway, so I'm not sure why this would matter. If anything it's more accurate.

View count is used to guide to price embedded sponsors, so in the short term matters while things get recalibrated.

What if its both? ;)

How many?

Only YouTube can tell, that's the fun part.

And because only YouTube knows this, they can tell us anything they want.