If you don't specify the column list then this implicitly assumes a column list including all user insert-able columns (non
computed) as below.
CREATE TABLE #T
C INT IDENTITY,
INSERT INTO #T
This is far more generally useful as the most common reason for having an
IDENTITY column is to allow SQL Server to manage the values.
I suppose it could imply that if there was an extra column in the insert source on an
identity table that it should generate an execution plan for the explicit
identity_insert case but not much benefit to this IMO.