From a technical perspective this question is easy to answer. Here's why:
Referential integrity over distinct databases is something which - as of my knowledge - is not defined as part of any SQL standard. Meaning you will have to enforce such a "feature" on your own.
In practical terms this means the following: let's suppose you load some data into different tables residing on different database servers. These two tables may have a logical relationship (kind of a "logical foreign key") named ORDER_NO (which is the "primary key" on server s1 and the "foreign key" on server s2). Now in order to ensure "logical referential integrity" between these two tables, your application has to make sure that the "master" record (on server s1) is written and committed before the related "detail" records are written and committed to server s2.
And the same holds true for TDM: as long as you make sure (via your TDM plans) that the table on s1 is loaded and committed before the table on s2 is written, everything is fine.
Again, as of my understanding. It's up to you to make sure that those tables are masked in the correct order, then these referential constraints will remain intact.