Acella: non coincident parcel address problem

This page is to describe and discuss what a potential problem with EAS addresses in the adoption of the Acella system.

Definition

We'll call this the non coincident parcel address problem.
Some EAS addresses contain parcels that are not spatially coincident with their address point.
A picture is worth a thousand words.

This is the parcel footprint of "1011 Silliman St" which includes parcels 5935-027 and 5935-028.
Parcel  5935-028 is not spatially coincident with the address point (in green).

Here is the same address with imagery. This looks like a single home.

You can see this for yourself in EAS.
Acella does not support this sort of address.
In Acella, parcels for an address must be spatially coincident with the address point.

Variation

There is quite a bit of variation to these data.

This one has 3 contiguous parcels.

This one is looks complicated but is not - it's just linked to 2 over lapping retired parcels, plus another unretired parcel... OK, that's a little bit complicated.

 
This one is non-contiguous and seems likely to be an error of some sort. 

Universe

Here is a spreadsheet that contains all of these addresses.
You can see these in detail in EAS.
There are 3253 unretired parcel level addresses currently being used at DBI. 
This is out of some 300,000 plus addresses. 
We can find these easily with a query like this.

SELECT
	ab.base_address_num,
	ab.base_address_suffix,
	a.unit_num,
	sn.base_street_name,
	sn.street_type,
	p.blk_lot
FROM address_base ab
inner join addresses a on (ab.address_base_id = a.address_base_id)
inner join streetnames sn on (ab.street_segment_id = sn.street_segment_id)
inner join street_segments ss on (ab.street_segment_id = ss.street_segment_id)
left outer join address_x_parcels axp on (axp.address_id = a.address_id)
left outer join parcels p on (axp.parcel_id = p.parcel_id)
WHERE 1 = 1
and sn.category = 'MAP'
and ab.retire_tms is null
and a.retire_tms is null
and axp.retire_tms is null
and not st_intersects(ab.geometry, p.geometry)
order by
	ss.st_name,
	ss.st_type,
	ab.base_address_num,
	ab.base_address_suffix,
	a.unit_num;

Discussion

We were planning to adapt the existing EAS change notification process to feed address changes into the Acella GIS. 
In doing so we would not have to stick with the exact model layed out in the XML file.
In fact, it may help to think about the Acella address data as a view (in the MVC sense) of the EAS address....
and perhaps the address point in Acella GIS is not the address base point but rather the parcel centroid.

Another possibility is for EAS to adopt a muli-point geometry for the base address.

Another other ideas?

Questions

Does DBI have to support this kind of address?

Can we live with the Acella constraint of allowing only parcels to be part of an address if the parcel is spatially coincident?