I manage many servers with >=384GB RAM. They maybe detect an error once a year, with several never reporting an error in its life. If I noticed a server reporting errors more than twice a year I'd probably swap its memory, and if that didn't fix the problem I'd probably replace the server.