From what I can tell, it's all about the right mindset and what configuration reinforces it. I don't see alarms as either good nor bad per say, just like wet switches vs no wet switches. My personal opinion at this point is that it depends on the overall configuration and perhaps on the individual. I'm tempted to say there is one right way but i've changed my mind enough times to know better.

My latest attempt at certainty (which is subject to change without notice):

For an alarm to be valuable, the diver and the alarm have to be effective enough to overcome the added risks that seem to come with a greater sense of reliability. The added attention gaps in monitoring are virtually inevitable when you think your risk has diminished. the diver has to adequately resist the temptation to let down their guard. Further, there may be degrees of this that are involuntary, in other words no matter what the training or mental efforts are, some people, maybe all, may not be able to fully resist this temptation, at least not all the time. the question then is: what are the risks that the alarm will go unnoticed all together after lulling a person into relying on it, or that a gap in monitoring will coincide with a malfunction in the "alarm/monitoring system" resulting in no alarm? either could be deadly.

I'm tempted to conclude that you are at greater risk from spacing out when the risk is intermittent than from spacing out when you know the risk of life threatening consequences is constant. One is unpredictable, the other is less so.

I'd venture to say that your monitoring interval is proportional to the perception of consequences and that that has the overarching influence on life sustaining outcome. No matter how full proof your system is, never "trust" it and no matter how experienced you get, never get too confident.

this has lead me, in part, to go from the Evolution with state of the art monitoring and alarms, to a mCCR Shearwater Copis, with no alarms but with the sentient monitoring system between my ears fully engaged. I don't think artificial intelligence is as reliable as real intelligence, at least not yet. Am I right? Who knows, and since a relatively few people have died on all rebreathers I might go a lifetime without being one of the unlucky, even if there is a flaw in my configuration or habits. But I have been so baffled by the near 0% (1) fatality rate on mCCR's vs the 150 or so on eCCR's that I've started focussing my attention mostly on good mindset and the kit configuration that I believe best reinforces it, not the perfect set of electronics.

In any case, I feel less confident of my current rig, yet ultimately less at risk of a error or malfunction going unnoticed.