Since releasing HL2, Valve has reinvented one franchise (Team Fortress) and launched two new ones: Portal and Left 4 Dead. Portal's inheritance from HL2 is relatively obvious. Not only was it packaged with Half-Life 2 and its episodic follow-ups, but it exists within the same world and is filled with recognizable sounds, objects, concepts, and mechanics. Team Fortress is only tangentially related--it started life as a Quake mod but was adapted to the HL world via Team Fortress Classic, and it has been a flagship in Valve's business model of entertainment-as-a-service. By contrast, the Left 4 Dead franchise is a brilliant culmination of concepts that were introduced in HL2, improved in L4D, and improved again in L4D2.
First, there's the basic gameplay paradigm: you are a survivor getting help from other survivors while trying to make your way through an urban setting filled with zombies. This is straight out of the "Ravenholm" chapter of HL2, where you traverse city streets and rooftops killing zombies with the aid of a NPC survivor. Ravenholm (and some later chapters) even feature proto-versions of the L4D panic events, where you have to dig in and defend against a wave of zombie attacks. Later in HL2 your role changes from individual-on-a-mission to squad-leader, which again feels very similar to the single-player gameplay of L4D.
The two games also have remarkably similar level-design, both visually and conceptually. Both take place in ruined civilizations. Both try to tell a lot of back-story using visual information worked into the scenery, including heavy use of graffiti. Both consist of linear paths that ape open-worlds. And both use a lot of the same "tricks" to draw the player towards the correct path.
Now, Valve has done a great deal of work to streamline their production without sacrificing game quality. The use of the AI director to spawn enemies in the L4D games, as well as new ways to play test that didn't eat into as much time for as many developers, meant that they were able to make a larger, more complex sequel to L4D in single year after spending 3 years on the original. And the ease of development here is a good indication of why Valve is moving away from monolithic story-puzzlers like HL and HL2 and towards smaller, iterable games like Portal or the Half-Life episodes or games that are more geared towards replay like TF2 and the L4D series.
For perspective, the L4D games both have only about 4 or 5 hours of gameplay if played once through. But they're still considered "full" games because they're meant to be replayed, and you can have dramatically different experiences from one play-through to the next. Half-Life 2 has closer to 12 hours of gameplay, all of it linear, and much of it puzzle-based. Geographically, it's substantially larger: any one of its chapters may be the same size as a full L4D campaign, and the vehicle levels are simply huge by comparison. And Half-Life is thematically about isolation, so you end up with setpieces that take place in huge, open spaces. L4D is more thematically tied to a claustrophobic feeling, so you end up with smaller areas. With the L4D games, you make your map, have the director plant enemies, and play test. In Half-Life games, you make 5 times as many larger maps, you place enemies manually, then you playtest your combat levels, readjust enemy placement, playtest your puzzles, readjust the maps, readjust enemy placements for adjusted maps, playtest combat with new enemy placements, and so forth.
When you add in the idea that early levels are often tutorials for later levels with more complex puzzles, you're looking at a huge development process. And if you're dedicated to quality, then you're going to end up with a long development cycle. Small wonder, then, that Half-Life 2 took five years to develop, and why we're seeing 4-5 hour "episodes" instead of a third full game.
That said, I've found the Half-Life games quite rewarding, if not as replayable, and I'm looking forward to starting the episodes where they add commentary tracks and (in episode 2) achievements. And then I'll be in the rabble clamoring for episode 3, no doubt.