After reviewing the various hexcrawl techniques I have access to, here are my conclusions.
Mixed Dice Speeds Up Table Look-Ups: This was extremely obvious from the wilderness hexplore example. By using one type of die for each of two or three stages in the process, you can roll a bunch of dice at once to quickly get a result. In contrast, rolling percentile dice for every table, as with the experiment with Central Casting, means you have to do each table one at a time.
Split Concepts for More Variety: Although Champions of ZED and Central Casting didn't exploit this as much as they could, splitting elevation and climate/vegetation into two rolls allows adding more terrain types without adding more table entries.
Broad Areas Speed Up Exploration: Both the Champions of ZED and the way I handled Central Casting involved determining one or two factors for an entire region at once, instead of hex by hex. Hex by hex rolls for everything slows things down.
Rolls Relative to Previous Rolls Maintain Logic: To a certain extent, the DMG does this. However, it really stood out in Victor Raymond's system, where the die roll shifts terrain type one step up or down or randomly selects between adjacent hexes. That makes it easy to build custom terrain lists for just one region, but use the exact same procedure for every list.
Reserve Details Until Needed: Both Alexey's system and Jed's system defer rolls for more details until they are actually relevant; you don't roll for details of a village, for example, unless the players choose to approach it. All they know is "you see a village near the hills to the west". This cuts down the number of rolls a little more, speeding up exploration.
The Broad Areas principle and the Reserve Details principle together mean that hex-by-hex rolls are best reserved for discovering tiny details. In other words, you roll per hex to see if there is a ruined tower or a hermit's hut, not for the terrain in a hex.