This is not really the case. I have personally quite literally implemented papers the week after they appeared on ArXiv and they worked. Anecdotical yes, and there have been papers that nobody was able to replicate, but at large the community is quite good at recognising which work can be used, and this work is then implemented and disseminated.