According to Ad Manager program policies "When displaying Google ads on their Sites, Partner must also comply with the AdSense Program Policies, wherein 'partner' is referred to as 'publisher,'" and Adsense rules are clear on auto-refresh. [youtube.com...]

Even if it's allowed, I'm not sure if the second or third ad would be as valuable as first one, or if it would drive your rpm down due smart pricing.

If you go ahead with it, you could run a couple of A/B tests using Adsense channels for your scenarios.

@Levo, thanks for the video. It is clear.
What I describe in my original post is allowed, because every case it would be as a result of a user interaction.

Even if it's allowed, I'm not sure if the second or third ad would be as valuable as first one, or if it would drive your rpm down due smart pricing.

I'm not sure I get your logic. If I had created "traditional" webpage (PHP style) that refreshed each time the user filled in and submitted the form, there would be no question regarding the value of the ads. Each submit would result in a new ad and a very annoyed user. With the SPA the user no longer has to wait for the pages to refresh. The user can and does interact seamlessly with the page. Why is then less valuable to show a new ad? To the contrary interaction with the page, is proof of engagement and interest in the topic. Two thing that I would imagine advertisers would value.

What does smart pricing have to do with it?

This makes me think of another point. The probability that an ad will be clicked is most likely a function of the time or duration that an ad remains in visible. In other words the longer the ad is in front of a user the higher the probability that it will be clicked. But if a user hasn't clicked an ad yet, probability that the user will in the future likely diminishes with time. In another words if user has no intention of clicking an ad no amount of time will change that. So there is a trade-off, refresh too soon will result in fewer clicks or wait to too long and you'll miss your chance to try again with a better suited ad. A/B testing may be able to find the optimal point, but I doubt I have sufficient traffic to come to statistically significant result in a reasonable amount of time.

Yeah, I'm not sure on smart pricing part, or how the refreshed ads would be priced (bidding-wise). My traditional way of thinking tells me that it's like a page with multiple ads and in my experience less ads drives the competition/value up. Each ad would also get a fraction of Average Viewable Time.

Each ad would also get a fraction of Average Viewable Time.

What isn't clear to me is if there is any tangible value in this metric.

Active view viewable (AVV), is tangible in that an impression is deemed viewable or it is not, and there are published reports that show that AVV is used as bidding metric. But even AVV, is questionable, the general rule of thumb states that one should strive for >50%, fine, but I doubt that any one could present evidence that increasing AVV from 50 to 55% would make any noticeable difference in revenue.

According to the AdSense docs on viewability:

An ad is counted as viewable if at least 50% of its area was displayed on-screen for at least one second (the minimum criteria according to Interactive Advertising Bureau (IAB) standards).

So now go back to AVT, 1 second gets you to the AVV threshold and with only 50% AVV you "get a benefit", so does having an AVT 10 vs 12 second really make a difference?

Furthermore, the metric appear in the AdSense dashboard, but I was unable to find it AdManager reports. AVV is there but not AVT?

This goes back to my closing point of my previous post, there certainly is value in showing an ad for a longer duration but there is a threshold at which point there is not benefit. AVT is certainly metric to help measure this.

I did a bit of research on this, and what I have found is that there is very little research available, most of it dates from around 2016 or 2017 around the time that AdSense rolled out its viewability metrics. What I have found is that duration is measured in the order of tens of seconds.

This report suggests,10 seconds:

“You can feature a striking visual or brand logo at the top where even if it’s not a full 100% or even 75% in-view, makes an impact if the ad is rendered for 10 seconds,” Manatt said.

source: [adexchanger.com...]

another, talks about 14 seconds:

On average, to be looked at for up to a second an ad needs to be viewable for 14 seconds.

It further shows to be "looked at" 4 seconds the ad must be viewable for 37 seconds.
source: [medium.com...]

Another blog post I saw reported that AdSense's Average Viewable Time is measured only up to 3 minutes, but there was no reference for that assertion. I haven't found any mention in Google's docs.

It would appear from the above that 1 minute would likely be a reasonable cut-off.

Some more info here [support.google.com...]

Ad Manager requires publishers to declare auto-refreshed ads so advertisers (or bidding algos) could take that into account. That might suggest they see them as less valuable?

I still don't think you're allowed to auto-refresh Adsense ads, and you can also leave it as is; a placement that displays the highest-bidding ad longer. [ezoic.com...]

The Google support link is very interesting, I had seen it before and was aware of the declaration requirement. But I hadn't seen the "Available minimum intervals for refresh" table.

User action-based == no minimum.
Event-triggered & Time-based == 30 seconds minimum / 60 seconds recommended

The ezoic link is hell-a-confusing, and makes no sense, it's starts with CPM's then talk about click through rates of 5% and then 1% and then halving the 1%. Clearly adding more ads to a page has a cannibalizing effect. I don't think that any reasonable publishers expects that if he/she is earning 2$ rpm with one ad that adding a second will double that to 4$. But the second ad will likely increase the page RPM from $2 to say $2.25, as a publisher one is still earning $0.25 cents more, so who cares if you were earning 2$ per impression rpm per ad unit before and now your are earning $1.125, you are putting more money in your pocket and that is ultimately what counts.

And clearly this cannibalization effect will increase with more ads (beyond simply diminishing returns), so there will be an optimal point, an optimal number of ads to show.

This dynamic holds true for refreshing ads as well.