I agree that trying to establish bright-line rules often seems a futile exercise because it's so easy to come up with ethical inconsistencies. I do have a theory about the moral calculus of conflict but it's inchoate and this comment isn't really the place for it. To sketch the outline, though, the severity of a given outcome in a conflict can be mitigated by the degree of risk involved in bringing it about; so while the imposition involved in sleep deprivation or keeping someone cold is mild compared to someone being horribly maimed by a piece of shrapnel, it's the massive prior asymmetry between captor and prisoner that's problematic in the case of torture. Of course, you could argue that this asymmetry is also strong where only one side in a conflict has an air force or where one side is using remote-controlled military assets like drones, whose 'pilots' experience no physical risks whatsoever. This is going to become a huge problem within a few short years with robotic infantry assets - as soon as they're 'good enough' to deploy effectively there will be huge pressure to do so, even if they are just slightly extending forward positions over a short distance.

Yes. Waterboarding Khalid Sheikh Mohammed is viewed by the vast majority to be morally far much worse than bombing a munitions factory. That is why doing it costs the tremendous political capital.

The deaths caused by the aggressor of a war are not consistent with international conventions, they are, in the words of a judge in at the Nuremberg tribunals, the consequence of the ultimate war crime.