Based on the papers you cite, it seems like its not even straightforward in U.S. law. They discuss three main legal questions that are obstacles to implementing an open-access policy:
- Is there a "teacher exception" to work for hire, giving researchers initial copyright over their papers?
- Does a non-exclusive partial rights transfer survive a subsequent exclusive copyright transfer?
- Is a university policy a sufficient legal instrument to achieve the non-exclusive rights transfer in question (2)?
Turns out, this scenario is specific enough that it really matters which jurisdiction is being considered. I took this question as an opportunity to research the copyright regimes of France, Germany and Japan (translations of their respective copyright acts linked). Japanese copyright law is fairly similar to U.S. copyright law in terms of rights transfer, so I will mainly be analyzing the first two (that and I have no competence in Japanese).
I am not in any position to be writing papers over this subject, but in considering questions (2) and (3) with a civil law context, I'll cover some laws on governing rights transfers. I'm not considering question (1) because if the university holds original copyright, then it is trivial for them to implement an open-access policy. I thus assume the faculty members hold original copyright for this answer (which is generally the case in France and Germany anyways).
Author's rights basics
Author's rights are separated into two branches: moral rights and economic rights.
Moral rights are generally non-transferable. Often they cannot be waived and last for eternity. Economic rights are those which can be transferred and exploited, but as you note these may also be subject to restrictions. For the rest of this answer, I will be avoiding the term "copyright" as it is ambiguous: It can mean just the economic rights (like in the translated Japanese), or author's right as a whole (like in the translated German and French).
Note: Links from this point on are in French (English resources weren't sufficient).
Survival of non-exclusive rights transfers
In Germany, this is a straightforward affirmative: Section 33 states:
Exclusive and non-exclusive rights of use shall remain effective with respect to rights of use granted later. [...]
In France, an answer to this question is elusive. This isn't too surprising as French statute hardly references non-exclusive licences. As an example, it took until 2007 for the GPL to be recognized in court. Given that the courts have been leaning towards giving weight to open licenses, my assumption is that they would follow the common-sense approach taken by German law though I've not found any direct statement to that effect.
Implementing an open-access university policy
From an author's right perspective, the biggest issue I see in drafting such a policy is that by default the authors give the university a non-exclusive right of distribution for future articles.
France has particularly strong protections for future works. L131-1 states:
Total transfer of future works shall be null and void.
Though it may look as if this can be easily avoided by adding a few simple small clauses, jurisprudence has been to interpret this in favour of the author when possible. For further information, see here.
While I'm of the opinion the scope of the policy would be narrow enough to avoid the reach of L131-1, there are additional restrictions on publication contracts. Noting that giving the university non-exclusive distribution rights will likely make the university a "publisher" in the eyes of the law, L132-4 states that:
A clause by which the author undertakes to afford a right of preference to a publisher for the publication of his future works of clearly specified kinds shall be lawful.
Such right shall be limited, for each kind of work, to five new works as from the day of signature of the publishing contract concluded for the first work or to works produced by the author within a period of five years from that same date.
This makes it difficult to have a blanket open-access policy. I'm not certain whether an opt-out clause would be enough to avoid the above restriction.
Additionally, France has the moral right of retraction allowing the author to withdraw granted rights of use under strict conditions (L121-4). French moral rights can't be waived so it's futile to account for it in a policy, but it's something to be aware of when implementing an open-access system.
In Germany, while there are some protections for future unknown types of use, the scope is fairly well defined here. This puts the situation squarely under Section 40:
(1) A contract in which the author undertakes to grant rights of use in future works which are not specified in any way or are only referred to by type shall be made in writing. The contract may be terminated by either party after a period of five years following its conclusion. The term of notice shall be six months, unless a shorter term is agreed.
I therefore don't see an issue with this in Germany provided that the policy is specifically agreed to and is renewed with faculty at least every 5 years.
Whether or not the university implements a lawful open-access policy (which is challenging in France...), the author has another available option. Both Germany (Section 38(4)) and France (L533-4 I. of the Research Code) have legislated a limited form of open-access that the author has a right to. While the laws are slightly different, they boil down to the following:
After publication in a journal, the author may publish the article in an open-access manner after an embargo period of at most 12 months notwithstanding any exclusive rights transfer to a publisher, provided that the research was at least half funded by public funds.