At the AV Club, Todd VanDerWerff and Noel Murray have an “inventory” on a subject that is near and dear to my heart: failed backdoor pilots (or as I prefer to call them, “stealth pilots”) from successful shows.
You know the kind of episode they mean; a show presents an episode that is mostly about characters we’ve never seen before, because this episode is going to be shown to the network as a pilot for a proposed series. When the series gets picked up, it’s still mildly weird to find an episode of All in the Family where Archie disappears early on, and we spend the rest of the episode getting to know the Jeffersons’ new neighbours. But it becomes a truly surreal and confusing experience when the pilot was rejected, as most pilots are. That’s probably one reason why the practice became less common as time went on. The reason for making a backdoor pilot is that it defrays the cost: instead of needing to raise the money to make a separate pilot, it’s covered by the budget of the existing series, and if the pilot fails, its costs are recouped in syndication. But “who the hell are these people?” episodes can hurt the show’s value in syndication, which loses the production company much of what it gained, financially, from the backdoor-pilot process.
One thing you’ll notice if you look at the episode guides for these shows is that the backdoor pilots were usually made toward the very end of the season. Sometimes, as with “His Two Right Arms” from Mary Tyler Moore, it was actually the last episode of the season. Remember, the concept of the “season finale” as a special event didn’t really exist until recently; combine that with the natural tendency of shows to have weaker episodes later in the production cycle (because everybody’s so tired), and it made some sense to have a backdoor pilot late in the season: it gives the regular actors some rest at a point when they’re tired out, and it also allows the burned-out staff writers to regroup or move on to something else. (In the case of “His Two Right Arms,” the backdoor pilot was actually done by writers who had no connection with the parent show. Other times it’s written by a creator who isn’t normally writing scripts for the show.) There’s another reason why this kind of pilot is harder to do now: the end of the season is no longer a creative/ratings dead spot, so prime-time real estate is too valuable to waste on a bunch of unknown characters.
The comments on the article also mention lots of other failed backdoors; for example, the next-to-last episode of Green Acres was a bad backdoor pilot, but so was the last episode, a bad pilot for a proposed series about Elaine Joyce as a wacky secretary and Richard Deacon as her long-suffering boss. There are a surprising number of shows where the very last episode is not a finale but an attempted pilot — because the series finale wasn’t a big deal either, and it was more important for the producers to beg the network to keep them all employed.