Those miners wouldn't even have a way to know it is a solution to a block.
Well that is the point isn't it?
Clearly this would need to be flushed out and tested and instaban is simply rhetoric. I'm not going to engineer the entire test here in a forum post, but that's the top level pseudo code. However any miner that fails the test should be considered suspect and perhaps subject to an additional test. A miner with a 1% error rate would have a 1 in 1,000 chance of artificially failing 2 known good blocks. Make it 3 tests and the odds raise to 1 in 100,000
You are still not getting it. Some ASIC chips don't attempt every value in nonce range. Some never do under any conditions, the multcore chips break the nonce range into "chunks" with a chunk going to each core. They never have 100% of the cores active because nothing has 100% yield, some are turned on. So if the nonce that solves the block is 7289372 and it is sent to core 2 but core 2 is disabled then that miners will NEVER return a solution for that work.
Of course you ignored the rest, how long do you wait? how much do you want to lose to orphans because you are waiting to publish a winning block and every second you delay the probability of another pool broadcasting a block and orphanning you increases. What about a miners that due to latency returns the solution after you publish the winning block?
It isn't a viable system.
Well, like I said, a forum post does not a system make.
This is a problem that has to be solved, or pooled mining will die. We aren't going to solve (or at least I'm not going to try) in this thread, but a sanity check of hardware is not an unheard of thing, and it's clearly needed now.
Folks are not going to stick around while they watch 20% of their earnings evaporate.