mapfinder has been running to find a suitable map that pops an SGL for about 4 days now, so most of my play has been the first 1000 years before abandoning lately. But a few lessons have been interesting lately.
The first game was conquest on a huge warlord map. These were leftover maps from earlier, abandoned runs. Several maps have all suffered from the same deficiency as the emperor conquest game that went to domination instead - missing opponents. And the map is so big that it takes forever to definitely prove that one is missing at all, much less mobilize to invade an island. So that hasn't worked out despite a lot of play time.
But my thinking on efficiency shifted a lot in the latest game. Previously "efficiency" meant that the units I lost did not die in vain and we never let a spear heal. MW's traveled in larger stacks and never began an assault that they couldn't finish. That's a fine strategy for more traditional play styles (especially on higher levels), but on speed games for the HOF I've started thinking of mounted warrior turns the way I think of worker turns. Every single turn must accomplish something, and an injured unit is almost a good thing in the same way a busy worker is a good thing. His idleness is accomplishing something by healing from the last fight. A healthy and idle MW is unacceptable, and if I have two MW's ahead of a stack of eight, the two will proceed past a enemy position to a more distant objective only for the sake of getting them further away from home. In sense, each tile further from home is an investment in that unit. generally lead units do whatever they can to be helpful and get some damage so they can wait on others to catch up. Sometimes that's a risky assault on a useless town, sometimes it's pillaging key tiles, sometimes it's taking out isolated units in the open. Maybe that's simple and obvious, but I've never seen that philosophy discussed and it really helped. My first raiding party was 3 MW's who had nothing better to do than start a fight and got to heal between the first town and reinforcements.
Victory conditions have started to break out into two start categories in my mind - Production (20k, conquest) and expansion (science, histographic, domination, and 100k). The conquest condition especially weirds me out because you've got so many competing priorities for workers and growth and units that it's hard to know what to prioritize. It feels like a 8 dimensional drag race where each turn is crucial and you must complete some key tasks like worker builds but that they really set a city back on progress to making MW's. 20k is a lot easier of course since there's only 1 goal.
But that expansion category tends to be a relatively straightforward process of initial . Right now I'm working on the chieftain 100k victory condition on a standard map trying to take the number 1 slot. In previous games like the histographic game with the Iroquois I've muddled along with just 1 settler factory, which paradoxically is kind of required on 60% maps selected for dom limit. But on my 100k maps mapfinder can run purely for 3 cow starts with a river, and 3 settler factory starts can be a selection criteria. I've probably played around 40 maps and only gotten 1 SGL on a map that I abandoned, but the start process has mostly become:
1. Settle capital in location allowing at least 2 other river towns access to a cow.
2. Build warrior while town grows. Water and road grassland cows first. research Masonry at full speed.
3. When warrior completes switch to settler unless food is short. That could be because I moved away from cows in step 1, or because cows are on plains. Send warrior exploring the immediate area trying to find additional resources and terrain to support setter factories. Abandon if 3 factories are not possible, even before discovering Masonry. The SGL is great, but it doesn't change the map.
4. If SGL build pyramids.
5. If no SGL but the map has 4 settler factory potential I'll save it and build settlers until I have 5 or 6 towns, 4 for settlers and 2 for ToA and pyramids, then switch to granaries.
6. If at any time a warrior is built because of production/food balance issue I don't feel bad at all. They research half the ancient age techs for me. And I generally lose no opportunity to declare on a backwards AI, take their worker if possible, and pillage key tiles before asking for peace. They're no threat obviously, but it doesn't hurt to set them back 20 turns. Side note: is there anything worse than popping philosophy from a hut with 1 turn left on a tech? Is there anything you can do to prevent this?
One game has proceeded to 1550 BC based on the 4 factory criteria with 4 grassland cows, a wheat, and 2 forest game, plus 10 Bonus Grasslands in the radius of the 4 relevant cities.
One of those cities could not go on a river so the wheat is required to boost food to 5, and will be irrigated to allow a forest citizen. Right now 3 of those are 4 turning settlers with some tedious tile swapping involved. The wheat city was the last to be founded and is prebuilding the ToA. This is nothing special but it's good practice while waiting for an SGL.
Currently I'm playing with the Sumerians, but am running mapfinder for the celts. Had a discussion on Spoonwoods HOF thread about it, and it looks to me like the advantages of each are:
Celts: Shorter anarchy (saves ~4 turns on finish?), 1/2 price cathedrals more easily whipped than full price
Sumerians: more likely SGL's (made up by patience with celts), no GA with pyramids, and earlier cheap libraries (but a big shield gap to whip the cathedrals)
The ultra early golden age sucks with the Celts, but I think it can be effectively used by getting settler factories working earlier and catching up on infrastructure before the end of the GA. Bonus grass tiles suddenly throw off 2 shields, and if you've irrigated the cows already the factories are pretty easy to run. If you get 2 pumps working initially it's pretty huge to be producing settlers every 2 turns from 3000 BC.
The biggest difficulty in all this is the need for a fast ToA prebuild. If there are n settler factories then I initially made my n+1 and n+2 towns wonder builds, but one could easily argue that the 2nd level wonder town should go in first to collect shields on ToA since that is critical path to earning culture. An even more radical approach with 4 factories is to use a factory location to build the wonder before converting it to settlers. The more I think about it the more right that last answer is, but I haven't done the math on it. In fact, the nearby cities should probably make workers to add into the ToA city after improving key territory. Unfortunately in this situation 03 is not on a river, so it won't grow past 6, and the ToA will be a bit slow.
The capital can do it with 4 citizens - 2 cows, FP and the wheat with a mine, which gets 10 food and 7 shields with the last 3 shields coming from the hill on growth. And you're also right that the second city can get the 10th shield from growth, either by reducing the pop to 5 and mining the wheat or growing onto the forest. 
