You can start to see some permanent deformation at about a delta T of 250 F.
The trick is to maximize the difference in temperature between the area being heated and the adjacent areas. When possible, I clamp an angle or a bar to "outline" the area being heated. I use water soaked rags on absorb the heat that defuses beyond the bar (or angle) and to maximize the delta T.
Heating starts in the most restrained location, at the apex of the triangle as described by eekpod, and I move the torch toward the free edge using a zig-zag pattern. Do not heat the entire area uniformly. Heat the area directly under the torch until it is to temperature and move on along the zig-zag trace. In the case of the beam, the flange is the free edge, and it too has to be heated, but not until the web has been heated and cooled. You can air cool it or you can use water soaked rags to hasten the cooling as long as you didn't heat it above the lower temperature of transformation which is 1330 F for carbon steel. There is no reason to heat it that hot, good results can be achieved at temperatures of about 800 F without damaging the steel member. In the case of the flange, move the torch across the width of the flange starting at one edge and moving steadily toward the opposite edge. Move an inch or two down the length of the flange and heat again. The bigger the torch, the better. Heat the area quickly as possible with the torch and then allow it to cool.
Now for the reality, if you are working with a plain carbon steel (ASTM A36), even if you heat it above 1330F and cool it with wet rags or an air spray and water, there is too little carbon to cause problems with excessive hardness and loss of ductility. The same can't be said as the carbon content increases or the base metal is quench and tempered. In the case of quench and tempered base metal, you don't want to exceed the tempering temperature. The trouble is, most welders do not know the carbon content of the base metal they are working with. If they don't know the carbon content and the other alloying elements, they'll not be able to calculate carbon equivalency. So, to be safety the "rule" is to keep the temperature below the lower transformation temperature and the operation shouldn't cause any unacceptable microstructure if the base metal is a plain carbon steel or a high strength low alloy steel.
Best regards - Al