Choosing the cheapest optimizer cost

Right now, the optimizer chooses the path with the cheapest cost.

However, do we take into account the behavior of the plan in handling
mis-estimated row counts? For example, if a path has a log(n) behavior
for changes in the row count, and another plan that is slightly cheaper
has a log(n^2) behavior, should we choose the former, knowing the the
row counts are often inaccurate?

I suppose one approach would be to track not only the path costs, but
the handling of mis-estimated, and account for that in the final path
choice? Do we already do this by giving less stable plans higher costs?
Does that have the same effect?

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ As you are, so once was I. As I am, so you will be. +
+ Ancient Roman grave inscription +