EDIT: The proposal is actually for a runoff election in place of a hand recount, in which the top-two vote earners advance to a second, independent, contest. I misread the original article. These particular arguments against in terms of voting strategy don't necessarily apply anymore, but I still hold quite firmly that our voting laws have served us quite well during this pretty rare occurrence, even if the hand-recount has been annoying. Spending a ton of extra money to exclude a legitimately filed candidate in the name of decisive victory is still stupid.
Two Minnesota state legislators are proposing instant runoff elections as an alternative to recounts to decide super-close state-wide races. The legislative sponsors of this measure emphasize the increased perception of openness, clarity and trust in an electorate-centric runoff election as opposed to a bureaucrat-centric recount.
While this article cites only the high cost of a second, complete election (they put the tab at "at least a few million dollars"), IRV is more than just financially impractical.
Basically, instant runoff elections work by a system of ranking. Using the U.S. Senate election as an example, each individual voter would rank each of the candidates (we'll use Franken, Coleman, and Barkley for the purposes of this exercise) either 1, 2, or 3, 1 being the most favorable. Then, whichever candidate receives the fewest first-place votes in the first round of voting is eliminated, and whichever of the two remaining candidates has the most first place votes (with the third candidate removed) wins the election. Thus, every time the winning candidate will win with more than 50% of the vote.
There are two reasons in particular that this method seems more desirable than a plurality-wins election. First, it increases the likelihood that individuals will vote their actual preference--especially if that is a third-party candidate--because even if their preferred candidate gets very few votes, their vote can still "count;" that is, their preference is not eliminated from the actual calculus of the race just because their vote went to a candidate who received very little support.
Second, it allows us to vote for more than one candidate if more than one candidate is palatable to them. Suppose you approve of both Al Franken and Dean Barkley, but hate Norm Coleman. Your vote is the "anyone but Coleman" vote, and you could potentially vote for both of the candidates you support, doubling your chances of avoiding the candidate you dislike.
Assuming that the end goal of the democratic process is representation that best reflects the desires of the community, this method, on its face, appears to limit strategic voting, and allow individual voters to offer their electoral support to each candidate they find representative of their will.
In reality, not so much.
The ranking process of IRV has its own set of screwy outcomes. Because it gives second-choice candidates the same weight no matter the preference, several voters may end up offering their votes to candidates they do not actually support. If you were a Barkley voter who hated both Franken and Coleman, for example, you would end up supporting your second-choice Franken just as would a gung-ho Franken voter, or a Barkley voter who also liked Franken. Perhaps this could be fixed by making the ranking system optional, but then your vote for a third party is often eliminated anyway, reinforcing the same strategic voting issues as plurality-wins.
Furthermore, suppose most of the electorate placed either Coleman or Franken in first place, but also approved of Barkley, placing him second. Barkley still loses in the first round because so few put him first, even though the majority of voters thought Barkley could represent their views reasonably well.
And maybe somewhat more cynically, even though Minnesota has the highest voter turnout in the nation, still, just about 77% chose to express one preference. The education costs to create an accurate ranked preference might be significant, and this model works best assuming that individual voters have a clear preference and can articulate it.
This is not to say that IRV is on some level worse than a plurality-wins model, only that it trades in old issues for new issues, and costs a whole heck of a lot to implement in the way these legislators have suggested. Over all, it seems that a more clear and streamlined election process--in either case--would ensure that more individuals are better able to express their preferences in the first place.