Hmm, how about survival long enough to reproduce within the context of local conditions? Of course, if the environment changes (globally or locally) that may suddenly change the criteria for fitness, right?
Do you think selection just eliminates individuals at random?
Yes, it can, as there are multiple dangers out there for organisms--e.g. the newly born/hatched of 'most' species ;-) have not yet evolved resistance to teeth or digestive juices of their predators. That's why these things happen "on average" isn't it?
To quote an earlier post of mine on an unknown thread, what if there is a squirrel that has the 1.0 edition of new, improved paws for climbing faster to evade predators, but it never passes on its genes for such because it ran under the tires of my car before mating? Such an event would have an effect on the observed rate of adaptation... :-)
Selection is statistical. Allele change in a population is gradual. Consider the casino. Does the house always win? No. Does the house make a profit? Yes.
Life plays with an unlimited budget, because the sun keeps pumping the money in. The populations that have the statistical edge will usually prosper. But things change, sometimes suddenly, and extinctions occur.