I think that a key issue here could be reliability / availability of the system. Safety engineering commonly relies upon 'layered' safety systems, with ultimate safety not being dependent upon any single system UNLESS that system has a very high level of safety integrity. If you have the latter, you may be able to remove one or more of the 'back-up' safety measures. Sprinklers are statistically a very reliable life safety measure (less so as regards asset protection) - and extensive records of satisfactory historical performance is an entirely valid means of proving high integrity.
So far as I'm aware, we do not have the same historical or analytical evidence of safety integrity for oxygen reduction / nitrogen concentration systems. Therefore, speaking personally, I'd not be happy to consider them as a primary safety system (e.g. to support significant deviation from good practice as regards means of escape - like sprinklers) - I'd want to see some demonstrable 'high-integrity' performance as a secondary system (e.g. used successfully as asset protection) first.
If you were really determined to use it as a primary safety system, some sort of FMECA (Failure Modes, Effects and Criticality Analysis) might be appropriate? You'd need some system reliability/availability data from the supplier, though.
In essence, until I was pretty sure that the system would work 99.9999% of the time, I wouldn't be using its presence to get rid of too many of the 'tried and trusted' good practice risk reduction measures that have have years worth of proven effectiveness behind them.