Notice
This is a detailed topic, and is not discussed on the list in such a way that entire posts will shed light.  So I have summarized, and included heavily edited posts from different individuals to illustrate some points.

Rebreathers have dozens of ways to kill you.  Some have more ways than others, and there have been many heated debates over this fact.  For now, this topic will just reproduce some of the different failure modes that have been discussed.

First, I'll differentiate between the three major classes of rebreathers, fully closed electronic, fully closed non-electronic (O2, manual add), and semi-closed

First, we look at fully closed electronic, systems like the U.S. military Mk 15,16,15.5, Cis Lunar Mk 5, Buddy Inspiration, BMR 500, etc. . .
The basic failure modes for electronic rebreathers are:  (in addition to what all rebreathers have)

Next we have fully closed, non-electronic units, systems like the Lar5 O2 rebreather.   These are inherently simple rebreathers, but are depth limited.  All O2 rebreathers have few extra problems above what normal rebreathers have (see below)

Next we have semi-closed units.  Thses units have automatic addition of gas through orifices (Drager Atlantis/Dolphin), passive addition

All rebreathers can suffer from the following failure modes:

single sensor failure

This happens, but as Rod Farb notes:

Date:    Wed, 15 Jan 1997 14:39:24 -0500 (EST)
From:    Roderick Farb <rfarb@email.unc.edu>
Subject: Re: D.E.M.A. SHOW

Three is the number of sensors to use in life support. Sensor arrays work and work well.   In my applications of the CCR, I have sensors and use them but I do not have to have them to fly the unit. If they die, you can dive the unit semi-closed until you run out of gas.

So, what happens when one sensor fails and you are using three?  Not much.
First you have to notice a problem.
After this, check all three sensors to figure out what is going on.
Then decide what to do.  Quite often, at this point it is a non-issue.  sensors die.  That is why electronic rebreathers have three. This post from Dude shows how much alarm there is over one sensor failing.

Date: Thu, 16 Jan 1997 13:28:46 -0800
From: heyydude@pipeline.com (Mr. Dude)
Subject: Re: Electronic vs. Mechanical

At 35 minutes into the dive at 140, I looked at my Primary, and it was showing me an "Alarm" plus "O". I knew that this meant that I had a sensor that was out of range, so I immediately scanned all three sensors with my Secondary. Sensor 1 was at 1.2 (my set-point), Sensor 2 was somewhere between .8 and .9, Sensor 3 was at 1.2

My solenoid did not fire to add additional O2 as would be indicated by your description of the electronics of the 155. In fact, I continued to dive with the solenoid firing only when Sensors 1 & 3 indicated lower ppO2 than the set-point of 1.2 Once they reached 1.2, the solenoid stopped firing.

Also, another strange thing happened that I had NO explanation for : As I continued the dive, the "Alarm" light on the Primary display began to dim. "This is weird" I thought to myself. Eventually, it dimmed to the point where it went out.

I scanned the three sensors on my Secondary, and now Sensor #2 was reading 1.2 again.

I called Dick the next day and asked him about this - he said that indeed, the "Alarm" light would fade up/down depending upon the condition of the bad sensor, and that the sensor could "drift" like this.

This is a "feature" of the 155 that I had never heard of, but I thought it was pretty cool anyway.

Kevin.

Given that a single sensor failure can and does occur, what happens?  It depends on the design of the electronics.  Rod gives a nice (albeit long) summary for some of the rebreathers he's familiar with.

Date: Thu, 16 Jan 1997 12:15:09 -0500 (EST)
From: Roderick Farb rfarb@email.unc.edu
Subject: Re: Electronic vs. Mechanical  

What are the ramifications of losing one sensor and what do you do about it? When one sensor goes out, the solenoid begins firing adding more O2 to the breathing bag. If you let this go to completion- until the solenoid quits firing- then the breathing bag will contain higher pO2 amount than was preset. Why does this happen? The way the electronics work in the 15, 15.5, CCR1000, 500, BUT NOT THE 16 is this: The pO2 from each of the three sensors are summed and the result divided by three- by this it determines the average pO2 of the three sensors and this average value must be within 5% of the preset value or the solenoid is ordered to add O2 to the breathing bag until the preset O2 is reached again.

The problem is the the electronics ALWAYS DIVIDES the sum of all sensor pO2 values by three even if there are only two sensors reporting. Since the sum of two sensors is less than the sum of three sensors and since the divisor is always three, then in the two sensor case, the result is a low pO2 value being reported which initiates the command to the solenoid to fire in more O2 to the breathing bag until the preset value is reached. This is what it would mean on a dive using a unit that was preset to deliver a constant pO2 of 1.2. With all three sensors working properly, they each report a pO2 of 1.2. The sum of all three reports is 3.6, and 3.6 divided by three gives 1.2, the average. Everything is fine and no O2 need be added to the breathing bag. Now, suddenly you lose a sensor. Two sensors report pO2 of 1.2. The sum of the two reports is 2.4 which when DIVIDED BY 3 is 0.8.  This triggers the solenoid to add O2 to breathing bag until it perceives that the 1.2 pO2 setpoint (preset) is achieved. But the pO2 in the breathing bag will be much higher than the setpoint because in order for the electronics to believe that there is pO2 of 1.2 in breathing bag each of the two remaining sensors must report a pO2 of 1.8 in the breathing bag. 1.8 times 2 equals 3.6. And 3.6 divided by 3 equals 1.2. So, if you lose one sensor as in this example, the solenoid will quit adding oxygen when the pO2 in the breathing bag is 1.8 because the electronics thinks it is 1.2. If you don't understand this, don't buy the rebreather.

What do you do if one sensor goes out ( you hear the solenoid fire more frequently when this happens and you will see on your primary gauge the 0 light on and green and the A (alarm) light on and red. If this happens, reach around and turn off the electronics. Continue diving the unit manually referring to the secondary gauge for pO2 information. When pO2 in breathing bag falls below where you want it, add a spurt of O2 manually.   Add the O2 slowly. The solenoid only adds 200 ml at a time because the sensors take time to react to the added O2 and you don't want to over shoot the value you want. If you do overshoot, breathe the unit semi-closed by exhaling through your nose and add diluent. Then start adding O2 again. If this is too scary for you, then abort the dive and replace the sensor. However, the units can be dived manually all day long without hassle, since it takes a little time (depending on work load) to drop pO2 in the breathing loop to where you would need to add O2. The Mk 16 electronics sums the pO2 reported by two of the three sensors that are the closest together in value. In other words, it uses voting logic whereby it drops the third sensor value whose pO2 report is off the greatest from the other two. It sums the two and divides by two to get the average.

Rod

In other words, to properly deal with a single sensor failure, you must be aware of the design of the electronics.  This also underscores the theory that you should be able to manually check each sensor individually so that you can use your brain instead of the electronics to get you home alive.  A single sensor failure does not necessarily mean you abort the dive, however, it does mean you have to be vigilant from that point on.

Now, you want the real kicker?  It turns out that Rod was wrong on one issue, and it was brought to his attention.  His response:

Date: Thu, 16 Jan 1997 23:27:39 -0500 (EST)
From: Roderick Farb <rfarb@email.unc.edu>
Subject: Re: Electronic vs. Mechanical

Barrie, you are right on the money. The Mk15.5 has a clipping circuit. Any sensor that deviates more the 43% from setpoint has its information discarded and an electronic limiting signal is used in its place so it doesn't significantly change the average of the other two. I always thought it went to zero. Can you check for me on the 0.66 of setpoint for the limiting signal for the divergent sensor. For me it means that my bailout for the loss of one sensor was ultraconservative. If I run at setpoint of 1.3 and the limiting signal is o.66 of setpoint then the loss of one sensor would elevate the pO2 in the bag to 1.47 which is no biggie to me since I would like to run the unit at 1.4 but can't get the unit to hold it. Now what I can do is calibrate the unit for 1.4 and disconnect one sensor. Sweet.

multiple sensor failure

How can this happen?  Easy.

Date: 16 Jan 97 19:14:41 EST
From: "Mike (CA)" <102023.2362@CompuServe.COM>
Subject: Re: D.E.M.A. SHOW

So let's go over some what-ifs.
3)sensor failure (all).
    this would only happen if the scrubber was fully flooded, would require going to semi-closed or oc mode.
So you can see there are a number of solutions to sensor problems, and the fact that I know what my ppo2 is at all times, is going to give me the information I need for deco and o2 tox. I'm not comfortable with not knowing what's in the bag on assent or desent.

Michael Filloon
JUST ADD WATER SPORTS
SANTA MONICA, CA

What to do?

Date:    Fri, 17 Jan 1997 16:00:09 -0500 (EST)
From:    Roderick Farb <rfarb@email.unc.edu>
Subject: Re: BioMarine O2 sensors

If two sensors fail simultaneously, turn off the electronics, ascend monitoring the single remaining sensor if you believe it is giving correct info. Say sensor one is alive and the others are dead. If sensor has been accurate during the dive, there is no reason to believe it will be aberrant because the other two failed. You gotta manually fly the unit. If you are concerned about sensor one or just want to play it safe, breathe the unit semi-closed and use diluent supply. Rod

complete electronics failure

Date: 16 Jan 97 19:14:41 EST
From: "Mike (CA)" <102023.2362@CompuServe.COM>
Subject: Re: D.E.M.A. SHOW

So let's go over some what-ifs.
1) The elecronic flood or the battery craps out.
    simply monitor the secondary display and manually add O2.

Michael Filloon
JUST ADD WATER SPORTS
SANTA MONICA, CA

Summary

Electronics have some unique failure modes.  The biggest danger is that you may not notice the problem until it is too late.  This is why you hear phrases like "Silent Death" or "Upon recovery the diver is dead, and the rebreather is fine".  It is not always obvious what can go wrong, and it can be difficult to determine why someone dies on a rebreather.