There's "PPCoin", a fork that adds "Proof of Stake" alongside "Proof of Work" in an attempt to reduce the energy consumption of the mining process.
There's others as well. Some bitcoin proponents consider these "scam" coins as they have limited use other than speculation, others consider them experiments in different approaches. But they do show that forking bitcoin is fairly easy.
You wouldn't need to try and duplicate an ecosystem. People with nodes on the network would "vote with their CPU power" to support the version they liked the best - so the best one would naturally get the most support.
I don't think so. A "better" Bitcoin fork that has no MtGox, no BitInstant, no BitPay, no blockexplorer, etc. would probably have very low adoption.
You could harvest a large number of new chain coins before you go public, and offer an exchange service, but that would involve you owning alot of new-chain bitcoin, which would be a concern to people switing in.
