The "...could conceivably occur with GPS, Galileo, or BeiDou" part of the article isn't entirely true, though. Galileo is not operational (only four satellites have been launched, all proof-of-concepts), for starters. Beidou is a mixed constellation for which half the coverage doesn't have the access issues of a pure-MEO constellation. A GPS satellite could conceivably have the same problem, but it's easily corrected because GPS is supported by a network of ground stations with global coverage--corrected ephemerides can be uploaded virtually at a moment's notice.
The specific issue with GLONASS (bad ephemerides) would be easily corrected with a better ground network. As it is, you can only upload the appropriate data when they are over a limited, high-latitude portion of the earth. It's very much a single point-of-failure issue, which one can interpret as either a poor design decision or a by-product of pride / stubbornness (which, in engineering design, are frequently one and the same). It's a good reminder that, despite good scientific and technological underpinnings, the Russian space program still suffers from underdeveloped support and very limited global / international cooperation (not to mention testing and quality assurance issues).