There has only been one serious attempt to detect current life, the labeled release experiment on the Viking lander. It came back positive. If we are seriously interested in detecting current life we don't need a sample return. Any rover can dump some rock into a nutrient bath and put the results under a microscope. If anything is growing/moving, life is detected. Viking didn't have the bandwidth for such things. Lets do that a couple times first.
I mean, why not?
> Fossils would survive a million-year trip just fine.
Sure, but getting to pick where we take the rocks from helps us maximize the possibility of fossils in the sample. The rovers have found rocks with evidence of water - being able to pick those instead of a random chunk (which might've been blasted from deep bedrock) found in Antarctica has its appeal.
> There has only been one serious attempt to detect current life, the labeled release experiment on the Viking lander. It came back positive.
This seems like a good argument for sample return.
These particular samples will be from the present-day surface, with known geographic (aerographic?) origin, not affected by potentially millions of years of interplanetary travel, and not contaminated by sitting around on Earth for who knows how long.