I've been wondering for a while how to approach calculating Credit Valuation Adjustment (CVA) for a portfolio with embedded options. As a toy model take, say, a single contract (callable swap), no further collateral management or extinguisher. When following the usual ansatz (cf Brigo) trying to separate risk-free claim and cp-risk terms: AdjustedPV = RiskFreePV - OptDef_A + OptDef_Bwe want to know the option terms on the RHS. When valuing the swap cash flow we have to assume an excercise strategy: (i) excercise decision to maximise non-cp-risk adjusted time value (seems appropriate when I'm party C pricing a contingent-CDS on the A-B trade, since A will not considerthe credit quality of B in the excercise decision, if he is already hedged by my insurance).(ii) make excercise dependent on counterparty condition (makes sense if A is not hedged, and B is deteriorating rapidly. Then time values with/without CP risk are separating, wherethe former excercise strategy should have higher expected payofff than the latter).Probably the short answer is 'depends what you want to do with the value', but I was hoping for some more thoughts and experiences from others. From what I've seen at work and in the literature there is curiously never mentioning of optionality at all, yet there should be situations where (ii) is the one to use in modelling?