Tolerances do come into play in safety and security engineering, and also when designing for reliability. * What are the likely failure modes for this subsyste

Tolerances do come into play in safety and security engineering, and also when designing for reliability. * What are the likely failure modes for this subsystem? * What are the consequences if it fails? * How much should we invest in analysis and testing of this subsystem or aspect of the software? * When we find out about a severe issue internally (not known to the public), do we issue an "out of cycle" patch immediately, or do we wait for the next release? Do we tell customers about it? I know there's some moral hazard in our industry, because it is rare to see software companies advising users of security problems that weren't reported by a third party. The usual case (in closed-source) is that security issues discovered internally are quietly fixed without notifying the customer.