Lactose is just keeping a juvenile gene active and is a matter of life and death - the blonde gene is a single mutation that can take root from a genetic bottleneck. AMD even lactose isn't universal. Just being able to smell that soil is moist (instead of looking at it, or touching it...) is not as powerful a survival advantage as lactase or sickle cell or some of those life and death mutations for food or malaria resistance.
I think there is more information being transmitted than just that the soil is moist. The soil is lightly moist + highly porous + has high levels of microbiological activity which are the optimal conditions for seed germination. Consider that inedible seeds could act as a preserved food that could be germinated and edible in only a few days with no energy input by hunter gatherers who were known to gather and eat seeds - giving more like 70-80k years for Petrichor to be interpreted by humans as an attractive smell.
The plant growth/food hypothesis doesn't make sense to me because this signal is timing specific and the timing isn't related fruiting periods. Also growing plants have their own smells so why would an indirect smell be preferred?
Also, consider the evolutionary path of associating a smell doesn't require any protein production or chemical changes - only neurological ones. This shows that mice can inherit the ability fear certain smells from their parents. It's possible that it could happen epigenetically in humans as well considering the significant changes in taste and smell sense experienced in ones lifetime.
FWIW the lactase gene is not a matter of life or death — people lived at high latitude for millennia before it evolved. When it happened though it clearly was selected for. Most genes don’t select Individually on a life-or-death basis, especially in long-living species as its time scale is wrong (slight exception: pinch points like pandemics). It merely adds a bit of evolutionary advantage aggregated with all others. Thus, for example, your sickle cell example: it’s pervasive but hardly universal among its carrier population even though that population was until very recently all exposed to the same (malarial) risk.
In any case my point was that mutations certainly do evolve and stabilize within the span of human civilization (however one might define that). Evolution is a big, parallel, path-dependent process indifferent to which particular kinds of selection (physical or social, endemic or peculiar) happen to favor reproduction.
