Into the 1970's, NASA did that. That was why there was Viking 1 and 2, Voyager 1 and 2, Pioneer 10 and 11, etc. Since then, however, NASA has stopped doing that. It became a balancing act- yes, 0 to 1 is much more expensive than 1 to 2, (1 to n is not quite as cheap as it is with software but it's still much cheaper than 0->1), but NASA Science is in the business of answering questions. The question is, will building, launching, and operating (the expensive part) two Parker Solar Probe's and two Juno's answer more questions than building one Parker Solar Probe, one Juno, and one OSIRIS-Rex? Almost certainly the three different probes answers more questions than two copies of two different probes. So once launch vehicle reliability got to be good enough that the fear of total mission failure went down low enough (1), duplicate missions basically went away.
1: Edited to add: this is actually tied into the Space Shuttle in interesting ways. See T.A. Heppenheimer, _The Space Shuttle Decision_ for why the STS became the sole space launch system for all of the US Government. Of course if it's manned it's reliability has to be so high that you don't have to worry about loss of payload, so building two copies of it was no longer necessary.