The existing test is privately developed, opaque, and the responsible policy makers in the states using it, including California, can't get information on the validation from the maker. And, regardless of previous trends, which varied, every state using it saw a drop in 2017. This is all mentioned in the article.

Either there was a common national effect that dropped real outcomes, all states using the test independently made simultaneously missteps that dropped real outcomes, or the 2017 test was changed in some way which produced the downward movement.

If it weren't for the opacity, it would be easy to eliminate the last possibility.