For better it worse, the sophistication of scripting doesn't seem to matter: if you give $1M to the first non-coinbase tx in a block, eventually miners will write systems to take that money :(

If you write any protocol where you let miners choose the winner, you get this problem if the effect of winning is economically significant. Of course you shouldn't create protocols that stupid, but when there's a race to collect money from gullible people, engineering very much takes a back seat.

The real defence seems to be that economically useful actions on Bitcoin can outspend stupidity over the long term. I certainly hope this continues, and I suspect (though cannot prove!) enabling more people to use Bitcoin economically will further tip the balance in its favor.