Hate to break it to you, the terminals of CMOS battery are typically almost directly attached to a number of chips that are thousands of times more fragile than the battery itself.

Essentially, you can't fry a CMOS battery with a transient without also frying at least couple components that will prevent the motherboard from booting.

Is it possible for a motherboard to get CMOS battery fried without frying anything else vital? Maybe... but...

If you told me that you were in a roll over accident where you did not damage your car roof I would say "well, maybe, highly unlikely but sure, why not, maybe this happened to you". But if you said you were in 17 rollover accidents and in 15 of theme the roof was undamaged I would say "you are full of crap".

It is amazing that you're doubling down after getting downvoted into oblivion and then going off on car accidents? I'm surprised you aren't telling me that there is no way a bird could have fried a transformer... you know, because BIRDS AREN'T REAL. Remind me again, what motivation would I have to lie and make up a story about something like this?

Ok, I give up... you are totally right. All those machines just magically fixed themselves.

</facepalm>

I totally agree with twawaaay. There is a huge electrical separation between the incoming mains and the CMOS battery. It is very unlikely that a mains spike could damage the battery without also frying the intervening circuitry.

A more likely explanation is that the faulty servers had quietly deteriorated over the years, until a mains failure caused them to reboot, thus revealing that the batteries were faulty. And if the batteries had been sitting there for sometime with an internal short, then it is also likely that they would have visible damage (eg be leaking).

Also if they had been tested by simply pressing the reset button, they would likely boot just fine. However if they had the mains removed for a hour or so, it would be a different story..

I have been amateur EE and doing electronic repairs for fun for the past 2 decades.

When stuff fails due to external stimulus (like a lightning or connecting it the wrong way) things fail in a typical pattern starting from the power inlet, psu, and propagating into the device until it finds some things that can stop it. Voltage regulators, ripple capacitors, fuses, mosfet switches and so on are almost always dying first before any internal component.

Sometimes you get some stuff that is very fragile get hit somewhere in the middle of the circuit with everything around it intact. This is less likely than a typical path of destruction but happens.

But what you do not see is things that are quite tough get destroyed in a sea of a lot of fragile components that all somehow survive.

And that is because any device HAS to have a number of components on the power supply side that are put there with specific purpose of insulating the device from harmful outside signals as well as insulating outside from harmful signals from inside the device. It is not a choice of the designer, it is a regulatory requirement. You are required to do this to pass various certifications to be able to sell the device at all.

And so in all likelihood you (we) are right. The battery died on its own silently in the server while it was powered up. The battery does not do anything while the server is under power, it only takes over to power the internal RTC while there is no power supplied. Then you restart a number of servers and lo and behold, some of them have a boot warning possibly preventing the machine from rebooting complaining the battery is dead.

The servers were relatively new? It is possibly they were bought in bulk and all had the same faulty battery in them that had very short life span due to manufacturing issue. That isn't very rare.

This is way more likely than a lot of completely fine batteries get hit by a voltage spike in a middle of a fricking motherboard with nothing else damaged, in a large number of servers, the same way.

There it is! Appeal to Authority Fallacy.

What you're missing is that this is a totally different set of hardware and circumstances than you've ever experienced in your 2 decades.

Instead of coming out with a know-it-all it-cannot-be-possible attitude, it would be far better to approach things by asking questions or at least having an open mind. Innocent until proven guilty, right?

If anything, in those two decades, you'd have learned to ask basic questions. If you are truly interested in this, you could have asked me for the brand of mobo or to provide photo evidence of the batteries, mobos, etc... or even gone to the datacenter to see for yourself, but instead, you've just had this whole "i am right, you are wrong" attitude that really doesn't get anyone anywhere.

As I said before, I'm not lying. I have no reason to lie about this. It is what we experienced.

This is my last reply to you... waste of enough time.

The machines are not years old.

Plus, machines boot fine with a dead CMOS battery.

I can't even imagine a shorted CMOS battery preventing a boot.

I don't get it.

What is amazing is that you don't have to get it. Dead != Fried.

How does a dead battery prevent a reboot? The battery is isolated by a series diode to prevent it shunting the circuit. Certainly some parameters will be lost (eg time) but it should still boot.

Whatever, it's easy to test. Simply remove the battery from your PC and see if it will reboot. Then clip a lead across the (vacant) battery holder to see if a shorted battery prevents a reboot.

Well put, same goes for the sibling comment from the electronics hobbyist.