We can do a bit of guessing, but it is not about locking the view, it is about locking the row you are fetching for consistency purposes. So it will lock rows in all three databases if it fetches rows from them. It will also have database schema locks on db1, and it might have locks on the others to avoid you dropping the database (but it may not).
How long the data is locked it completely dependent on which isolation level you are in/any locking hints. In the default isolation level (read committed), you will generally lock one row at a time as it is fetched and released. Then the next row is fetched. Users can change rows you haven't fetched, and rows you have fetched.
In other isolation levels (like read uncommitted), no locks will be taken on the data, just the schema. Yet others (repeatable read and serializable) as data is fetched it will become locked and will stay locked. You are likely to get full table locks too.
Louis
Blog: http://sqlblog.com/blogs/louis_davidson
Book: http://drsql.org/ProSQLServerDatabaseDesign.aspx
