After a short introduction of multidimensional risk networks, we will present an example of risk network motivated by a company which must separate a "central manager branch" (CB) from other "subsidiary branches", necessary for coping with different types of catastrophes/claims. The subsidiaries must all be kept above 0 by capital injections subject to transaction costs, until the ruin of the central branch. The underlying question is how to split both the initial total capital and total income rate between the branches, in an optimal way.

Several problems, like for example the ruin probability with one subsidiary, admit exact answers as long as the CB is just a deterministic drift. On the other hand, non-deterministic CB's and several subsidiaries give rise to interesting approximation and asymptotics open questions, some of which will be discussed briefly. In particular, numeric results of dividend barrier optimization of a SNMAP approximation obtained by Pade approximation of the downward ladder Laplace transform will be presented.