A traditional stumbling block (though not the only one) is the question of religious schools. Voucher proposals have historically been seen as ways to get the government to fund religion, leading to various backlashes dating back to the 19th century: http://en.wikipedia.org/wiki/Blaine_Amendment
A solution could be to only allow vouchers to be used for secular schools, but that runs into opposition from religious conservatives, and it's also unclear if it'd be permissible to exclude all schools run by religious organizations from an otherwise generally available program. I suspect conservative support for school vouchers that allowed religious schools would take a huge nosedive as soon as Islamic schools started taking advantage of them though, which is currently what's causing the Netherlands to rethink its voucher program.
News flash. The government isn't funding anything. The government produces nothing and their only source of income is theft from productive individuals who do produce things.
We would end up with a worse two-tiered system than what we have now.
On the other hand, poor neighborhoods can be more dangerous and involve greater academic and social challenges. The solution is not to measure schools in poor areas against those in good ones and declare them worse because they perform more poorly (due to circumstances beyond teachers' control), but to measure students' baseline ability and then fund and reward relative improvement, rather than on the basis of absolute outcomes.
For example, say you go into a neighborhood on the first day of school and find that only 50% of 10th graders meet expectations for literacy. The best teacher in the world is not going to be be able to bring that up to the 95% level in a wealthy area on the other side of town, but if the proportion of students who are literate rises to 75% by the end of the 12th grade (correcting for dropout %ages), then that's a huge improvement. In economic terms, it's worth adding more funding right up to the point where marginal net gain falls below zero.
There are obviously willing and committed teachers willing to take on these important challenges. Maybe they would do better by setting up nonprofits and applying for funds to establish charter schools instead of abdicating their negotiation power to the national unions.
There is no system you can devise within the bounds of a capitalist society in which people with greater means don't achieve better educational outcomes.
I found the monopsony problem you've pointed out on the employment end of the system quite insightful. Entering an industry with a limited number of employers generally isn't a great idea. However, there's a simpler solution to the problem than more unionization, which in practice requires non-voluntary membership to be effective. I think we'd all be better off if the government continued to fund education for those who can't afford it, but stopped actually running the schools themselves. I don't understand why voucher proposals find such limited support.