This case is a good illustration of how our genealogy software doesn't serve us well. We focus on entering information into our software about people, when in actuality, the tasks we perform are searching for records about people, analyzing the information in the records, and recording what we have found.
If you have a Church of England parish register, generally you have an entry which describes a three-step process of the banns having been read over three subsequent weeks. I don't see any reason to break this out into three separate facts in a database. My choice would be the same as this answer, to use the date range to put the first and third dates, and enter what the source says in notes.
If you choose instead to enter a single date as in this answer, I would choose the third week because this is when the three-step process was completed. The third and final date is, by nature, closer in time to the record creation date.
Note that for Scotland, the page at CPR Banns & Marriages says:
Forthcoming marriages were supposed to be proclaimed on three
successive Sundays, however, in practice, all three proclamations
could be made on the same day on payment of a fee.
If you use a date range for the English parish registers which include three dates, and a single date for the marriages from Scotland where the actual event was only one day, as described above, you can see at a glance that the event which took place in Scotland is of a different nature.
I would only use three separate events if my sources were separate -- if I had entries about the reading of the banns that came from church minute books or a person's diary, where an individual reading was recorded on one particular Sunday.
Whichever method you choose -- especially if you choose to collapse all these different types of banns into a single recording method because it makes for a 'cleaner' display of the information in your software -- be consistent, and record your reasoning for that choice in your research notes.