arildno requested that I add this to the thread, as he didn't want to leave any gaps in his explanation:
Unfortunately, I glossed over a relevant topic because it is "too obvious", but on further reflection, I've found that my argument on why a viscous fluid favours downwash really becomes untenable without broaching it.
It concerns the "trivial" fact that for a viscous fluid, a stagnation point becomes locked onto the leading edge.
Clearly, that high pressure zone will give a fluid particle somewhat above the wing a horizontal acceleration component away from the leading stagnation point.
Thus, that fluid particle does not only, as I seemed to suggest, get a roughly normal acceleration onto the wing, but also a tangential acceleration down the wing, providing its "punch".
Locking the stagnation point on the leading edge effectively replaces the unphysical mechanism through which an ideal fluid effects tangential downrush:
It places its stagnation point on the downside of the wing, fluid rush up towards the leading edge, twists about, and rush downwards the upper side.
On the other side of the stagnation point on the underside, the fluid rush down to the trailing edge, twist about it, and the backflow then rush up to meet the downrush in a new stagnation point.
I.e, in the D'Alembert case, we have infinite suction pressure at BOTH edges..
Thus, the leading edge behavior in a real fluid is to replace a totally unphysical mechanism for downrush on the upper side with the mechanism of the frontal stagnation pressure, whereas viscosity's role at the trailing is to reduce upflow.
Both these mechanisms are succincntly described in russ waters' first link, i.e, that viscosity tends to DAMPEN velocity gradients.