It's interesting the different descriptions you come across for the effects of higher or lower cart loading. I own a Glider M2 (.8mv output, 24 ohm internal impedance, recommended loading "200 - 47K ohms"). I found the best choice for me to about 400 ohms. In my trials, lowering the loading value below about 5K made the sound progressively tighter, but changes above the several thousand ohms range made no significant further differences compared to 47K, sounding all basically the same, which is to say whispier and looser. Set the value too low, and things got constricted and mechanical sounding, with no 'air'. Set the value too high, and images became too diffuse, large, and see-through, with sloppy transient control and excessive 'air'. The right setting gave the best image focus vs. dynamic life, and precise transients combined with a natural sounding HF balance and LF forcefulness.
Curious about the recommendation, I talked to some folks more technically knowledgable than myself, who confirmed that above a few thousand ohms, the MC cart will basically be unloaded. My results tend to agree with this proposition, but if you look at Albert's results or the default loading on the Lucaschek of 22K, there would seem to be some disagreement about that. On the other hand, although Blackie's description of the sonic trends he observed coincides pretty well with my own, Marakanetz, Newbie, and Wlutke seem to be describing a situation where the sound gets 'softer' as they go lower, and 'harder' as they go higher, which I consider to be the opposite of what I found. But there might just be some confusion about language used here, because we can probably all agree that the extreme treble increases with increasing value.
It was my conclusion though, that tonal balance per se is not the most crucial quality being adjusted through loading optimization, but rather the degree of precision vs. freedom with with the stylus is allowed to follow the groove, vis-a-vis the electrical damping factor at work. This has its impact not only on the balance, but at the same time on imaging and transient behavior. I decided that if I could find the best loading for correct-sounding image focus, density, and size, with cleanly deliniated transients, but not so low as to make the sound small, stuffy, airless, and dynamically dead, then if at that point I still craved more top end 'air', I would try to address the balance elsewhere in the chain rather than raising the value and living with the resulting imprecision. Fortunately, or perhaps as a result, everything came together to sound accurately balanced at the chosen value, and I realized that my previous 47K loading, while enjoyable on some recordings, had not been a lifelike representation of the true HF balance or realistic image size.