For anyone who is not yet aware (and for those who are, simply skip to the next paragraph), the issue relates to a European patent application having a claim that is broader than anything in the priority application becoming invalid through lack of novelty once a divisional application is filed and published, because the divisional application becomes prior art under Article 54(3) EPC as a result of the claim not being entitled to priority. This strange result comes from a literal interpretation of the wording of Article 54(3), which makes no distinction between applications that may be related to the one in question and other applications (or indeed, as might be argued, the application itself).
My solution to the problem relates not to some unwritten rule that divisionals and priority documents should not be considered to be prior art (which some would argue), but to Article 88(2) EPC, which states:
"(2) Multiple priorities may be claimed in respect of a European patent application, notwithstanding the fact that they originated in different countries. Where appropriate, multiple priorities may be claimed for any one claim. Where multiple priorities are claimed, time limits which run from the date of priority shall run from the earliest date of priority".The simplest way of thinking about a claim having more than one priority would be where the claim contains alternatives, in which case it can be easily divided into two or more separate claims. Some alternatives might have a valid claim to priority, while others might not. Let's say a claim is directed to a widget made of copper or iron, and the priority document discloses only copper. The claim is entitled to priority only for the copper widget, but the claim can be rewritten anyway to: i) a widget made of copper; and ii) a widget made of iron. The first claim would be entitled to priority, and therefore not at risk if a divisional is filed, while the second lacks a valid priority claim but is novel over the priority disclosure, so would also not be at risk.
No distinction in Article 88(2) is made regarding the form the claim takes. More complicated forms of claims can therefore also have multiple claims to priority, although separating them out may be more tricky. To take a slightly more complicated example, let's say claim 1 is directed to a widget made of metal, and the priority document only discloses widgets made of copper. The claim cannot easily be divided into two from the wording alone, but can at least theoretically be considered to be two claims: i) a widget made of copper; and ii) a widget made of a metal other than copper. These claims together have the exact same scope as the original claim. As before, the first claim is entitled to priority and is therefore not at risk if a divisional is filed, while the second claim is novel over the priority disclosure and is therefore also not at risk.
Some of the discussions about how to tackle the conflicting divisionals problem focus on what kinds of disclaimers might be used to get around the problem in this kind of way. In my view this misses the point, although the idea does get part of the way there. The concept that a broad claim can theoretically be divided into two parts, one of which is entitled to priority and the other which is inevitably novel over the priority disclosure, is itself sufficient to overcome the problem even if the exact form of wording is not easily derived from the application. This is because the collective scope of the resulting two imaginary claims would be exactly the same as the original claim, and Article 88(2) EPC indicates that there is no need to actually rewrite the claim to be two separate claims. There is therefore no need to find the right form of wording to satisfy added matter requirements because no amendment is needed.
I see a nice parallel here with mathematics, where imaginary numbers are of great use in solving real problems, for example in finding the roots of a cubic function or in modelling electrical circuits in response to time-varying signals. Just because there is no real solution to the square root of -1 does not mean that an imaginary solution (usually termed i) cannot be a useful way of looking at a particular problem where this might arise. In the same way with problematic claims, just because there may not be a way of rewriting the claim as two parts, each of which would be valid for different reasons, does not mean that considering such imaginary claims cannot be a useful way of solving the problem of conflicting divisionals.
As usual I may, of course, be wrong. Any thoughts on the issue would be welcome.
Update 25 May: A commenter has pointed me to T 1222/11, which seems to support the idea. See point 11.5.5 in particular.
Update 15 June: There is an interesting discussion on the IPKat here on the related issue in the Nestec v Dualit case.
Update 31 November 2016 / 7 February 2017: I have been proved right! See my posts on the outcome of G 1/15 here and here.