In large Ruby deployments Schirp routinely patches out eval, patches out method_missing, freezes core classes, and forces production systems to be fully booted before serving traffic — to prevent third-party libraries (e.g. one that 30 minutes in production requires a random module and redefines the division operator) from silently corrupting the runtime. He argues Ruby developers shouldn't need to do this and would love a VM-level 'harden' method; the issue has been discussed on the Ruby issue tracker multiple times. Low-likelihood runtime capabilities × time = guarantee of problems on a thousand Ruby servers with 20 deploys/day.