Across the natural and social world, some fairly advanced and apparently robust systems destroy themselves without discernible reason, function, or purpose. Whereas such and similar cases have been discussed within specific branches of scholarships, the attempts to propose and test a general theory of self-destruction in complex systems remain, to our knowledge, a thing of the future. To move in this direction, we suggest that the conflicts between explicit stratification, thorough functional differentiation and tacit segmentation within complex systems, are likely to provoke their unintended self-annihilation. Furthermore, we hypothesize that whenever such systems collapse, it is the circular co-dependency of noisy signaling, redundancy, and semantic inflation that sets self-destructive mechanisms in motion. We illustrate our preliminary findings with examples from social, natural, and artificial life; reexamine autoimmunity as a popular model of systemic self-destruction; and indicate potential avenues of empirical research aimed at supporting, qualifying or disproving the underlying hypothesis.