If, say, the bears can't survive without a coat length of at least 10 cm, all bears with anything less than that will die.
(Emphasis mine)
But this simply doesn't happen, not over very short periods of time anyway. A bear with 9.5cm of hair will have a lower probability of survival than one with 10.0 cm of hair (or 9.6 cm of hair), but they won't all die. As well, there'll be areas of land where 9.3cm of hair is sufficent; the environment isn't homogenous.
Thus the genes for anything longer than 10 cm will survive, and above that number there will be variation. Genes will be unlikely to survive for hair beneath 10 cm.
Genetics tends not to work in such a clear cut fashion, it would take a long time to remove all short-hair genes from the gene pool, particularly if some are recessive.
If the following heat wave kills bears with a coat length of more than 8 cm, since all genetic information for such hair is gone - the DNA has lost the ability to create it - the bears will become extinct. Variation can only occur if the information for hair less than 8 cm exists.
A bear with 9.5cm would be more likely to survive than one with 11cm; this wouldn't happen overnight, and more likely the long-haired bear population would simply migrate into colder climbs (along with all the other flora and fauna that likes the colder conditions).
To gain such hair by a gain in mutation would be macroevolution. Such gains have not been observed, but even if they did happen the bears could not afford to wait until one came along.
Hardly, changes in hair length are non macro events, probably they can be achieved by relative simply genetic mutations, or even the simple recombination of existing genes. Note how almost every form of domesticated animal has long-haired varieties, and most have a fairly continous spread of hair lengths across different breeds.
™ Version 4.2
