Thanks for all of your replies. I think I will still analyse beta clauses at initial position as marked Themes, since in all the texts I've been analysing (in Spanish, across registers) this pattern does seem to signal discontinuities/shifts very clearly, which is very helpful in text analysis. Perhaps, as you seem to suggest, the thing is to decide how carefully you want to account for different layers of textual organisation and on which grounds (clause to clause within complexes...or higher-level waves, as in the analysis put forward by e.g. Martin 1992 and Martin & Rose 2007).I was just curious about that change in IFG! So in a way, perhaps, IFG's now sticking to (single) clause-wide patterns in a more strict way (without dismissing, neces[s]arily, the possibility of higher-level patterns, as the quote Hailing posted suggests)?
 A beta clause in a regressive sequence is not a marked Theme of a clause nexus; it is merely the Theme of a clause nexus. In order for it to be a marked Theme, there would have to be an unmarked Theme of a clause nexus with which it contrasts; there isn't. This theoretical misinterpretation can be sourced Martin (1992: 445ff).
 There is no change in IFG on the matter of beta clause as Theme in a clause nexus. The discussion has, instead, been given greater prominence in IFG4.
 The post from Hailing Yu included a quote (p551) from the discussion of beta clause as Theme in a clause nexus in IFG4 (pp549-53).