Here’s an example of Single Epoch RTK points that I collected at random times over several days.
The Green Sphere is to represent the size of a Golf Ball for comparison.
This is the precision you can get from a properly configured RTK System.
As you can see in these results, averaging 2-3 minutes of data for each one of these positions wouldn’t provide any benefit to each individual data point. The main reason ( I believe ) is that at this level of performance, the main sources of error aren’t random. They are timing and environmental. Performing more measurements under different conditions is what makes me start to trust the position, not collecting more data under the same conditions.
We typically consider a 6 hour mission to be the minimum.
In the past, I have used 1 month of PPP data to establish the position for a CORS.
Is that overkill, ya probably. But the Base is logging RAW data anyway 
It’s easy to submit data 2 weeks later when final orbits are available to feel confident in the Base Coordinates.
It’s important to distinguish the difference in precision and accuracy.
In GNSS, 1 answer is a guess. 2 or more independent answers start to give you some insight on precision. Calculating accuracy is a whole different problem, because we rarely know the “truth” without a LOT of work, or accepting someone else’s work.
Also, please don’t think that I’m disagreeing with your comments. I’m just offering other opinions as I enjoy the discussions 