I heard this anecdotally several times while writing CHECKDB and emergency mode repair for 2005. Spent several days trying to force SQL Server to make a database go suspect because of running out of space.

It's not possible in 2000 onwards. The transaction log reserves space in itself to guarantee that in-flight transactions can always rollback and not make the database go suspect. Hitting a corrupt page during a rollback will make the database go suspect however. Running out of space cannot, unless rollback hits a corrupt page (that it didn't hit during the actual transaction). That's corruption - that's not running out of space.

Thanks