I would see benefit in having a filtered INDEX as the target of an FK. I have run into such situations before.

For example, let’s say i have a system that has products, and imports information from another system that uses their own ids on a subset of products.

Product(MyId PK, Their_Id UNIQUE filter out NULLs);
Product_Info_From Other_System(Their_Id, info….)

I can modify the data on input to use our ids. However, if i want to keep the data as i get it, an FK to Their_Id would be nice. But, as they only have a subset of the products, i will have many NULLs in that COLUMN.

create table t1
(c1 int,
c2 int)
go
create unique index i1 on dbo.t1
(c1)
where c1 is not null
go
create table t2
(c1 int constraint fk_t2_t1 foreign key references t1(c1))
go

/*
Msg 1776, Level 16, State 0, Line 1
There are no primary or candidate keys in the referenced table ‘t1′ that match the referencing column list in the foreign key ‘fk_t2_t1′.
Msg 1750, Level 16, State 0, Line 1
Could not create constraint. See previous errors.
*/

I like the idea of being able to use that as a candidate index. I’ve submitted this as a connect item . The more people that vote for it the better the chance that they will implement it in the next release (or the one after that).