Bingo.
This is the root of the problem. The five-man rotation has evolved not because pitchers don't have the energy to pitch more innings, but in order to protect their arms.
The game developers have designed the game based on a flawed premise in contradiction to this reality. By linking innings thrown with energy/stamina, they're now backed into a corner where they've had to resort to a fix to prevent too many innings by starters in the playoffs.
This whole problem is the best example we have that the durability rating fails in reproducing accurate results because it's controlling two things at once - a player's "energy" as well as his tendency to get injured.
IMO, the best way to fix this is to have two ratings instead of one. Combined with a pitcher's stamina rating, it would solve the playoff issue without breaking the rotation scheme during the season.
Three ratings:
1) Durability - a player's tendency to get injured and his recovery rate.
2) Fatigue - a player's overall energy level. Would control the impact of playing/pitching time on his energy bar and govern attribute fluctuations accordingly as the durability rating does at present.
3) Stamina - how quickly a player loses his effectiveness in game.
Injuries would be controlled by #1 and #2 working in tandem. Continually pitching a fatigued player should have a detrimental affect, but more durable players can get away with it more.
So if a pitcher throws eight innings, his energy drops and recovers according to his fatigue rating.
A hidden, dynamic injury value would also drop and recover based on his durability rating and current energy level. Continually pitching a guy when he is fatigued would thus increase his chance of injury over the long haul.
Doing it this way would force you to protect arms during the course of the season to prevent injury and adjust rotations accordingly, which works better than the one-size-fits-all method of reducing innings by manipulating the energy bar directly.
The potential problem of overusing pitchers in the playoffs can be addressed in two ways: 1) Sever any link between fatigue and stamina. Instead, leave the stamina rating alone while reducing other attributes. (Or, at least reduce the impact of energy levels on stamina). And 2) for the cpu, adjust manager hook tendencies and implement a more realistic rotational strategy based on what others have already pointed out in this thread (extra bullpen arms, for example).
A separate fatigue rating may not be necessary if you think all players should recover from playing time at the same rate. This may be true for position players. I'm not sure about pitchers, though.
But the durability ratings influence should be on injuries, and nothing but.