Of course, GPU miner was not ready on day 1
Not knowable. The algorithm was discussed for a little while, and there was a testnet, development could have started before launch, and perhaps only a few tweaks were required at launch, if that. This is all speculation of course, but you can't say for sure how quickly a GPU miner was ready. Even to this day there are only private GPU miners.
Even the CPU miner was poorly written and poorly scrutinized (the whole design was rushed) so the public code was very unoptimized and still is not well optimized to this day. You could say the same about CryptoNight except that within a few weeks of launch, the Monero team was releasing optimized mining code (which got copied to the other CN coins) and now even the very best coders are struggling to make big gains.
So far, we have an account of a single entity at a ~2x advantage
You are way off. If you understood how GPUs work you would realize this.
Let's not forget the debacle surrounding monero's early days, which you call 'optimized' mining code
https://github.com/NoodleDoodleNoodleDoodleNoodleDoodleNoo/bitmonero/commit/3cc45e9324a402aee91e2f46861b2ca393d711aaWhich came out in may.wow, that was a shocker, how did that escape peer review by collective monero dev team? that was a huge red flag in my eyes & hardly a small thing to let slip through the net. Now that was much more worthy of the label 'instamine'
All coins are taking bits from each other. example
https://github.com/monero-project/bitmonero/commit/feac5a7b2ddce8f4612ecc459b176518061495b3credit to boolberry.
The 2x figure refers to efficiency & comes from here:
https://bitcointalk.org/index.php?topic=577267.msg7362060#msg7362060Unless either of us actually has wild-keccak GPU miner, I don't think we can say what the exact number is. The situation was the same with cryptonight.