Experienced FAH crunchers have observed that there are two different sets of tasks. One set is the non rigid tasks which has very small work units - takes 5 minutes on my computer. Then there are rigid tasks which takes just under an hour.

When the non-rigid tasks run, it brings the average FAH work time down. Then when my computer gets the rigid tasks, it bumps the estimated FAH work time back to normal, but bumps the estimated time of all other tasks to over a day to complete. Today, it almost went into panic mode - started running a series of high priority tasks. And took a long while for it to go back to normal.

Seriously if we knew ahead of time that there are two different sets of tasks and which one belongs in which, we could avoid this by having two different pools - rigid and non-rigid. Average time for non-rigid would be way down in the 5 minutes range and the rigid tasks would be 50 minutes (or whatever depending on your computer). If this can be done, then you would never have wild swings in estimated time.

Is there a technical reason why this could not be done?