It depends on what is burning and what the local winds are doing.
The good thing is, in the northern hemisphere the prevailing winds "most of the time" are out of the west, north west. Although on the coast you can get on shore winds, odds are more often than not the winds would carry most of the smoke and particulate matter out to sea.
As you say, sometimes the better part of valor is to abandon the problem and let it go. In this case though if the problem is super heated control rods in a dry storage pool, it is one of those things that will only get worse for along time, so it probably justifies some extraordinary efforts to control it like the folks at Chernobyl who did the initial helicopter drops to smother the reactor.
One of the issues is, that the Criticality of the fuel is determined not only by its amount, but in part by its geometry, and in part by how it is moderated.
I use the word Criticality to refer to how close it is to producing a self sustaining chain reaction.
http://en.wikipedia.org/wiki/Criticality_accidentOne of the ways they keep special nuclear material from going critical (starting a chain reaction) is to control its geometry. A given mass of material might be super critical (act like a bomb although very ineffeciently with only a few pounds explosive yield) if in the shape of a sphere. It might be critical ( acting like nuclear reactor with a self sustaining chain reaction) if it is shaped like a fat cylinder, and just below critical if shaped like a long thin rod. Same amount of material but the geometry changes how many neutrons actually hit another fissionable atom. In a sphere you have the maximum mass to surface area ratio, so relatively few neutrons escape the mass without running into another fissionable atom. In the fat cylinder you have more surface area for the same mass so more neutrons escape without causing another fission. As a thin rod, you have maximum surface area and most of the neutrons escape without hitting another fissionable atom.
If the fuel rods get hot enough to melt and run like candle wax they could puddle into a mass that is more sphere like than rod like and begin to produce lots of energy due to a chain reaction. That will create lots of heat usually followed by a small explosion that blows the molten mass apart, but in the process it would produce lots of new radioactive decay products and lots of small particles of highly radioactive material in the form of smoke. This is something that they very much want to avoid. I do not know for sure if the fuel elements chemical make up precludes that sort of melting scenario or not. It might be physically impossible for a serious criticality to occur and only a lesser situation where the metal gets very very hot and gives off very fine particulates (like smoke) that would spread large amounts of radioactive material about. I simply am not up on the specific fuel pellet design of these plants and what can or cannot happen at very high temperatures. In most cases I am familiar with the uranium is in the form of oxides so it cannot "burn" in the classic sense but it can get to very high temperatures.
Moderation can also cause a problem, as a mass that is in the wrong shape and just below a critical state (not quite capable of a chain reaction) can become critical (or super critical) if suddenly immersed in water. So they also have to worry about what happens if they let the fuel run dry and then try to flood it with water later.
Bottom line, this is the sort of situation that in most cases is better attacked vigorously and early. If you get behind the power curve it is like a spin on the salt at some point you cannot catch it any more, and your going for a ride no matter what you do. I don't think they are at that point of no return, but giving up too early could guarantee a bad situation would get worse. Unlike a house burning that you can back off and wait for it to burn out, this in my opinion would not be the sort of situation you would want to with draw from until you are absolutely sure you have no other option.
Someone with actual nuclear fuel design experience would have to address the details of what is possible and impossible due to the actual physical form and makeup of the fuel rods in this plant design. I don't have the background to address that specific question in detail only in theory of certain what if situations.
Some of you might find this article on the "Watt's Up with That" blog interesting:
http://wattsupwiththat.com/2011/03/17/live-real-time-monitoring-map-of-radiation-counts-in-the-usa/#more-36112Here are a couple other blog posts on this general topic:
http://wattsupwiththat.com/2011/03/17/in-light-of-radiation-fears-i-offer-this-repost/http://wattsupwiththat.com/2011/03/14/some-quotes-news-bytes-on-the-nuclear-energy-tsunami/http://wattsupwiththat.com/2011/03/11/nuclear-meltdown-race-to-save-reactors-in-japan/Larry