There are advantages to dying at age 30, too.

The Achilles' heel of such thinking is that you can find arguments for or against anything. Despite some workarounds that got selected for to make the "darn thing work at all", the general failure mode is quite evident from a phylogenetic trace, from an engineering standpoint and from the way that evolution on average only selects for neutral or greedily beneficial mutations.

After a certain point there was just no stochastical way for the retina to invert itself back to "sane architecture", it would have necessitated too many mutations at the same time. Same time, because otherwise the intermediate steps would be at a clear disadvantage and thus would not propagate.

chkno linked an excellent paper in the other response.