Very true, viscosity as a function of temperature isn't linear for a given base oil. Thus you cannot merely superimpose (mix) a number of different oils and expect them to display the properties of linearity as one would expect during linear superposition.
Certain base oils will thin much more quickly than others, while others will start out quite 'thin' comparatively-speaking, but will retain a much more steady viscosity over an entire temperature range. For example, GC and the XD-3 0W30; these oils are actually 'thicker' than many off-the-shelf 10W30's at operating temperature, yet qualify for the 0W30 grade because of their low-temperature properties.
Another way of looking at is by using an extreme example. If you mix heavy tar and sewing machine oil in your crankcase, the resulting mixture, although it might 'add' up mathematically to be an acceptable viscosity, definitely would not exhibit properties required for effective lubrication.
To answer the question of the topic: yes, mixing really does 'work', but oil ultimately is oil, and I think you will find relatively few in here who advocate mixing significantly different grades to achieve some specific results. 0W-30 is acceptable top-up oil to a fill of 5W-30. For example, a common 'mixture' recommended here is 10W30 and 15W-50 Mobil1 for LT1/LS1 engines -- the oils are practically identical to begin with, both 10W30 and 15W-50 are within the lubrication specs for the engine, and some people seem to have positive results with respect to valve noise with that combo. Fairly safe if you know what you're doing, but a disaster if you just haphazardly mix widely differing grades without an appreciation of their chemistrys, characteristics, or consequences.
[ September 22, 2004, 02:55 PM: Message edited by: pitzel ]