Then how about a group of developer upload encrypted code of the algorithm? They simply need to decrypt when it's needed.
That would still require trust in none of the developers leaking the selected algorithm.
True, but there aren't many choice in this case. Otherwise, the only choices are :
1. Picking acceptable PoW-algorithm that never seriously considered as SHA-256 alternative
2. Use flexible/configurable algorithm such as ProgPoW, but few parts of the algorithm/formula will be changed to prevent ASIC.
Maybe manufacturing the chips in China isn't really a problem. Many companies from other countries manufacture their products in China, and they control that the producers don't "tamper" with the design of their products. If the Chinese manufacturer manipulated the product, the customer would be able to charge a penalty for breaching the contract and change the manufacturer. An "open mining hardware" project could do the same thing, ensuring no Antbleed-style backdoors are implemented - and such a backdoor would be very easy to discover, above all if the firmware is open-source.
Once Bitcoin and the mining market becomes really big, it would be perhaps no problem to install production infrastructure outside of China able to deliver similar production costs.
Even if "open mining hardware" is popular and your scenario can't happen, i still think China still have someone advantage because :
1. Transportation cost is far lower if the chips is manufactured in China.
2. Faster transportation allows China to gain more edge since they're the first to use the chips in big scale.
But if mining on other country have maintenance/electricity price similar with China, then it's enough to prevent mining domination by china.