The "snowball clause" is yet another way that editors are encouraged to exercise common sense and avoid pointy, bureaucratic behavior. The snowball clause states:
If an issue does not have a of being accepted by a certain process, there's no need to run it through the entire process.
The snowball clause is designed to prevent editors from getting tangled up in long, mind-numbing, bureaucratic discussions over things that are foregone conclusions from the start. For example, if an article is speedily deleted for the wrong reason (not one of those listed in the criteria for speedy deletion), but doesn't have a snowball's chance in hell of surviving deletion through the normal article deletion process, there's no sense in resurrecting it and forcing everyone to go through the motions of deleting it yet again.
The snowball clause is not policy, and there are sometimes good reasons for pushing ahead against the flames anyway; well-aimed snowballs have, on rare occasions, made it through the inferno to reach their marks. The clause should be seen as a polite request not to waste everyone's time.
This test can be applied to an action only after it is performed, as the lack of snowballs in hell is not an absolute, and is thus useful for learning from experience.
The snowball clause may not always be appropriate if a particular outcome is merely "likely" or "quite likely", and there is a genuine and reasoned basis for disagreement. This is because discussions are not votes; it is important to be reasonably sure that there is little or no chance of accidentally excluding significant input or perspectives, or changing the weight of different views, if closed early. Especially, closers should beware of interpreting "early pile on" as necessarily showing how a discussion will end up. This can sometimes happen when a topic attracts high levels of attention from those engaged (or having a specific view) but slower attention from other less involved editors, perhaps with other points of view. It can sometimes be better to allow a few extra days even if current discussion seems very clearly to hold one opinion, to be sure that it really will be a snowball and as a courtesy to be sure that no significant input will be excluded if closed very soon. Cases like this are more about judgment than rules, however.