We introduce a family of risk networks composed from a) several subsidiary branches $U_i(t), i=1,...,I$ necessary for coping with different types of risks, which must all be kept above $0$, and b) a central branch (CB) which bails out the subsidiaries whenever necessary. Ruin occurs when the central branch is ruined. We find out that with one subsidiary ($I=1$), the finite time ruin probability of the central branch may be explicitly written out in terms of the finite time ruin probability of the subsidiary, provided that the CB in the absence of subsidiary bailouts is a deterministic drift. To study other problems, like for example the optimization of dividends to the CB with one subsidiary over a barrier, it is convenient to restrict to the case of phase-type claims to the subsidiary, and study the Markovian phase process at the moments when the CB process reaches new minima. The resulting structure is quite close to that of the phase of a PH/G/1 queue at the moments when it reaches new minima, and this yields in principle numeric approaches to several problems, based on the iterative calculation of the Laplace transforms of the upwards and downwards ladder times (also called excursions, and busy periods). In this work we propose a different approach of further providing matrix exponential approximations for the distributions of the ladder times. The advantage of this approach is that once a SNMAP approximation is obtained, many similar problems may be solved just by applying the recently developed scale matrix methodology. A numeric experiment for the CB barrier optimization problem is provided.