I don't agree at all with the argument that "is nullable" should not be set automatically for entities based on a view (even with the SQL issue mentioned).
Your same argument could be made that LLBLGen should not set the datatype of fields in entities based on views either, because if the datatype of a column in the underlying table changes, the view will not know about that either. Similarly, if they remove a column from the underlying table, the view won't know about that either, so maybe LLBLgen shouldn't create the fields either...
No - it makes sense for LLBLGen to do what it always does, which is make our life easier. Developers/DBAs should already know that if they change a table, they probably need to refresh the views and stored procedures/triggers that may reference it. And if they do that, they need to refresh the LLBLgen Catalog as well. Don't turn a DB server knowledge issue into an LLBLGen feature issue, just document the well known DB issue.
And b.t.w. forcing the developer to set the nullabilty manually doesn't solve the problem at all (the view will still go out of sync if the underlying table changes). Why not just LLBLGen do the work for us in the off chance that we all know what we are doing ;-) ?