The RF approach to safety was essentially deterministic. Any probabilistic treatment was limited to confirming that sequences outside the design basis (which was itself not comprehensively defined) were sufficiently unlikely (e.g. with an annual probability of less than 10-7). There seems to have been no overall integration of the diverse range of technologies covering nuclear propulsion, weapons systems, life support systems and operational systems, to cover the full spectrum of potential interactions between them. Instead, the strategy seemed to consider deliberately each area in isolation with a definition for each area of a worst-case accident that the other areas must withstand.
The engineering of the Kursk was similarly compartmentalized. This was possibly to minimize the need for detailed interface coverage between the various design bureaus. The flaw in this approach was that there could never have been a full recognition of the wide range of potential challenges, failure modes and consequences (including interactions) arising from internal plant failures and external hazards.