Do you like friction in your games...or is it a drag?

A recent post on TMP got me to thinking.  Where is the sweet spot between the eye in the sky wargame where everyone can always do exactly what you want and the version where units don't react to threats, move when their orders were to push on at all costs or simply do random weird sh....stuff?

That there should be some friction seems obvious to me the question is how much is enough?  Morale tests and reaction tests can go someway to creating doubt over outcomes.  But what about those events where a unit just can't move as fast as expected, or where ground that everyone thought was like a balling green turns out to be rough going, or troops who simply can't get the lead out?

I have toyed with various concepts over the years ranging from horribly complex but accurate, to gamey but workable and I'm still not sure where the sweet spot lies.  I know what I want just not how to achieve it.  The shopping list of desirable friction ideas should:

• Create unexpected variations in troop movement that are replicated for all units at the location
• Create delays in executing new orders linked to that units chain of command
• Create occasional refusals to follow orders based on the local commander's attitude to risk and reward and the situation in front of him.

I recall an interesting article on variable terrain effects from one of the wargaming magazines from years back that might fit the bill for the first concept.  An area or a linear feature is marked up as possibly being a variable feature and its true nature is diced for using a chart when the first unit reaches it.  So a river may be deeper than expected or shallower than expected, or a line of vegetation may actually be masking a stream or a ditch.  An area of ground may be boggy or rabbit holes are hidden under ground cover on a hill side.   Its a great idea but the charts need to be rooted in reality so no rabbit holes along the meadow fringing a lake and no unexpected rivers on hill sides the player's troops can see.

The other two ideas I'm beginning to think could be dealt with in a role playing style.  Give generals and sub generals personality traits. These might be linked to a chart that tells us how quickly he reacts (or even notices changes in orders or events affecting his command)  one chart  per general or just  a di modifier. linked to the personality trait.

What do you think are there any particularly good ideas out there that give consistant results and aren't too random?