• HL1 was an exploration of the possibility-space of having objects interact through emitted “energy” (i.e. by broadcasting messages that get attenuated by distance.)
• HL2 was an exploration of the possibility-space of putting a dynamics simulation engine in a game (and also spawned Gary’s Mod to continue said exploration.)
• The major offshoots of the series, Portal and its sequel, were explorations of the possibility-space created by combining “beams” of matter or energy (lasers, hitscan gunfire, tractor beams, paint), gravity, and wormholes.
In other words, each major new entry in the Half-Life universe was essentially a narrative wrapped around the results of playing around with implementing some journal paper (or reimplementing some prototype) of a novel type of realtime physical simulation, that had obvious game-mechanical implications.
I’m not surprised that we never got an HL3, because for the past 10-odd years, there hasn’t really been a big new obvious realtime physical-simulation tech to turn into game mechanics.
I’m also not surprised that Valve would return to making games under the Half-Life brand in order to explore the implications of a new technology (even if it’s not a new realtime physical-simulation technology. Or maybe they’ve got one of those up their sleeves as well...)