Agreed. Yet my field (Operations Research) has seen four or five new journals born in the past 10 years, all of relatively high level. Most often they were born as an initiative of one or more scholars in the field, who convinced a publisher that the journal would receive submissions.

So there is room for new challengers. However, now these scholars treat this journal as their precious pet project, would like everyone to submit articles to it and are (understandably) unwilling to embark in another, open-access initiative.

[edit: clarity]

It really depends on who the editors/founders are; e.g. the JMLR was automatically prestigious when founded by leaders of the ML community, forcing the previous crown bearer to become open access itself.