Good question. I found it easy because I've been through the pain of changing collations a couple of times (right to the point of rebuilding master to fix the server default collation).

Actually SQL Server is a a bit of a pain in its collation handling. Erland's Connect suggestion will go some way towards alleviating that pain (if MS takes it up - so everyone please vote for it) but only a small part of the way.

There is no good reason for a schemabound view to constrain the database default collation is it makes no use of it, its select statement uses no CTE that defines a column without an explicit collation being specified. Similarly for a check constraint, or a computed column, if no reference is made to the database default collation (so for example no string literals with default collation) there's no reason to object to a default collation change. So life could be made a little easier if MS would change the code to check whether the various objects are impacted by the default collation, and if none are allow a change of the default collation (perhaps with a warning, if MS thinks it's needed).

Then there's changing the collation of a column involved in one or more indexes: can we have an ALTER...COLLATION WITH REINDEX option? Having to drop the index and recreate it isn't too much of a pain, but it begins to get interesting when the column concerned is part of a key constraint, because if the column involved is part of a key there's the foreign key issue. Couldn't we have "on collation change cascade" as an option on foreign key constraint declaration? (OK, bad terminology, it's not cascading to lots of affected rows it's changing metadata -and that may need rebuilding some indexes.) Or is that going too far? Even without the foreign key issue there's the basic key issue - changing the collation may remove uniqueness - I need to be able to have the system tell me that my key will no longer be a key if I make the suggested collation change, not fine out the hard way (I recall that there's an issue with some digraphs and their single character equivalents for example is <thorn> equal to "th").

An interesting bunch of issues to be suggested by a QoTD.