Easy. To protect the environment and people. To make sure you don’t damage a neighbouring crop. To minimise the chances of the pesticide turning up in the wrong place and maximise the chances of keeping it on the market so that future crops can be protected. Some sensitive species are affected at very low doses, so off-target drift must be kept to the minimum. Plus, it is your legal obligation to confine the application to the treated area. Pesticides are very heavily assessed for environmental and human safety, but on the assumption that they are used properly, according to the label, and under conditions of ‘best practice’. What constitutes best practice is obviously subjective, but I know that the level of drift assumed by the regulator is not based on a flat fan nozzle in a howling gale with the boom flapping around your ears, and if it were, there would be even fewer products left for you to use. Finally, public perception influences policy and regulation and if a passer-by sees a drifty application, that reflects badly on the whole industry. There are good, easy-to-use, cost-effective tools for controlling drift, so there are no excuses!
All the time, except for when you shouldn’t! I firmly believe that some level of drift reduction should be the norm, and only switch back to conventional nozzles in exceptional circumstances. The ‘small droplet air-induction nozzle’ (which might give 25-50% drift reduction) I believe gives a good balance between drift and efficacy. The old HGCA nozzle chart (in sad need of updating) still gives good advice but I think we can now go further. Small weeds, and particularly grass weeds that don't capture much spray will be more effectively controlled with a conventional nozzle. Most fungicides can be applied with an AI nozzle. Conventional wisdom had it that insecticides had to have a fine spray to be effective, but the threat to chlorpyrifos suddenly had the manufacturers demonstrating (albeit rather too late) that it was effective with 75% drift reduction, so who knows!
In a perfect world, all products should be tested with drift-reducing nozzles so that we know exactly which ones work better with conventional nozzles, and we could do a proper cost-benefit analysis (both economic and environmental) of when it is right to use drift-reduction.