Many mutations are not caused by anything mutagenic - it is the replication system itself that allows for "errors" to creep in, as a major source of normal variation.

And "invisible" mutations, i.e. changes in the DNA strand during replication that do not impact any gene for anything phenotypic but occur in the 98% or some such of the organism's DNA that *doesn't* code for the organism, happen at a regular rate, and are part of the normal replication process, and thanks to THAT, we have a molecular clock and that is what can show how long ago two species, like us and the chimps (@5 mya) or the 'panzees and the bobobos (@3 mya), had their common ancestor.

If DNA replication were normally "error"-free we likely wouldn't have enough variation to speciate at all.

It's when a mutation affects *embryology* that you have a change that is visible to natural selection. At that point the change (possibly to a different species, possibly to nothing that works, e.g. extinction) starts to happen very rapidly, yet the basic rate of mutation is not changed.

"Natural selection has no goals" is not a meaningful statement, as natural selection is incapable of intention. It happens, and it is very much not random - selection selects something that works and selects against things that don't. And by work, I mean get more DNA into the next generation in such a way as to promote the survival and replication of the DNA *in* that next generation.

Remember the "goal" of DNA-based life is to become an *ancestor* - not just a parent.