The study shows that around half of supposedly "open" repositories are not technically open at all - their OAI-PMH endpoints are broken, outdated, or misconfigured:
https://doi.org/10.1177/01655515251396902
For me, this paper is not a revelation, but a confirmation of a simple point: open science does not start with policies or declarations. It starts with routine infrastructural work - making sure that the basic protocols actually work.
PDF accessibility in open repositories: A large-scale automated assessment