According to this NPR story:
http://www.npr.org/2014/03/31/...
Scott Oldham of Edmunds.com had a test drive of the Cobalt in 2004, with a GM engineer in the car. Multiple times Oldham's knee hit the key fob and car shut down.
Also, a major factor preventing identification of the ignition switch issue (or at least providing plausible deniability) is the part number. GM had 2 sets of cars: one set supposedly had this issue, the other did not. Both had the same ignition switch, so if there was a difference between the two sets, the ignition switch was not it.
Now we know the ignition switch was changed, but the part number stayed the same, making it difficult to correctly identify the issue. We're supposed to believe a single engineer was responsible for changing a part but not the part number?
Not that it matters much to me. My car searches start with Consumer Reports reviews and reliability ratings, and so no GM car has been in consideration for a while.