Perhaps some clarification of terminology is needed

The header of each register page refers to "Registration District and Sub-district" which is usually completed with two numbers separated by a dash or slash. There is a mapping between this 'Registration district' number and each borough/council (in modern parlance described as 'unit of local government').
So for example Smallburgh Rural District seems to map to 'Registration district/sub-district' 221/1 and 221/2. I've not seen any records for 221/1 or 221/2 which are not in Smallburgh RD, and nothing for Smallburgh RD which is not in 221/1 or 221/2. Yet

So having found someone on the 1939 Register you could say they were living in Smallburgh RD, or possibly 'Registration district/sub-district' 221/1, which is the more precise location (being the smallest unit systematically recorded on the page). Calling it a Registration District might be confusing, but that it what it says on the original document

As Mean_genie points out, there is a structure to the data below that level - people's names weren't just added to the Register in random order. The 1931 (and 1941?) census structure was apparently the primary basis.
Anecdotally, at least in rural areas, there is a mapping between register books (identified by a TNA Piece number) and individual parishes. Often one parish will be in one book, some larger parishes have two or more books, occasionally some books contain more than one (generally small) parish.
And that led to my question - has anyone attempted to build a database which gives the information like parish 'X' is in TNA piece "1234A" or TNA Piece "5678B" includes parish 'z'?
Chris, I was going to mention the FindMyPast maps in my original post. Sometimes they are very accurate which suggests somewhere at FindMyPast there is more detailed data that isn't displayed as part of each record. But as you say, other times the map just displays a general area. Where the original page doesn't give a placename I have used the online maps to work it out - but that doesn't work in reverse - for example if you want to find the Smith family who lived in Hickling.
The local history groups and one-place studies have probably found much of this information out already, and I suspect many other people have similarly found the correct parish/place name(s) for each piece. But what I haven't seen is any attempt to bring that information together in the form of an index similar to the way street indexes were produced before the census records were digitised and electronically searchable.