(a) It needs an absolutely ideal start position (lots of grassland). I wonder how many times he started a game, played a few turns and decided the start location was wrong. I'd bet it was a fair few, and these all (IMHO) count as ICS losing...
Depends on what you call a loss. For the amount of time it takes to milk a huge map to 2050 A.D. for a HoF game, I would not want to waste all that time (100+ hours) on a map that did not have the best start and cut my scoring potential a great deal. I had several maps that I played the first 40 turns or so and were still winnable, but I wouldn't have scored a whole lot of points (no luxuries nearby for example).
On MP any start in the middle of the jungle, desert, or plains with no river will be an automatic loss regardless of your play style unless you are playing a very poor player or they had an equally bad starting location.
- - - - - - - - - -
MAS -
MAS, obviously you've never witnessed the power of ICS. Maybe you are right and ICS will not win games in MP, we can only hope that is the case. I'd rather spend 100 shields building military units than spending that on a marketplace.
Building ICS you can mass-produce scouts and have a MINIMUM 20 of them out exploring and on a huge pangea map with only 8 civs you can easily get all the ancient era techs before 2500 B.C. without reloading. Difficulty level also affects what you get from huts. On Chieftain, just about every single hut will get you a tech or settler, while on Deity most will be empty.
Obviously MP won't be played on huge maps, but imagine if I you were building other units instead of scouts, as they wouldn't be really needed. The other thing about ICS is since the cities are so close to each other you don't need as many workers to get those cities to max production. If you also build roads, you could get the second city built on the same turn the settler was built, while the spaced city builder is spending 4 turns (or more) of having that settler walk to the city site. This 4 turns keeps getting compounded as time goes on, and the ICS builder would have 10 cities by the time you have 4. The 10 cities would each be having time in between the production of settlers to build military units to defend the city and all cities would be able to help defend each other. The spaced city builder needs some more reaction time to get more than 1 or 2 extra defenders at each of his big cities when an emergency comes up.
It won't be easy to attack the ICS player even if he has weak defenders. If he has only 3 warriors in a city, you end up using 3 units that turn taking just one of his many cities. Then your attackers (horseman, archers) are all open for a counterattack from all different directions. Actually, I think some players may leave a bunch of units outside of their cities (except the captial), so you can't take a peek at their garrison from the embassy spy option and know exactly how many units they have defending.
The only reason I think ICS will not survive against a human player is that the human is more likely to defend his capital with 5 veteran spearman instead of 2 or 3 regulars like the AI does, the human would be more prepared militarily and even possibly launch the first offensive. And a human would go for faster tech pace, but I doubt if they would reach Fuedalism (for pikeman), before the ICS player attacks. It would come down to who makes the first offensive and who gets the better start (terrain), IMO.