Ceramic SMD capacitor failure

[Pentium100]

I just finished repairing a server power supply. Finally tracked down the problem to a leaky ceramic cap (about 1uF) used for power-on reset. Its resistance was around 1M. I replaced it with a new one and now the power supply works OK (I hate 0603 components - they are a PITA to solder). Last time I encountered a bad non-electrolytic capacitor was in a >40 year old device (all the paper caps in it were leaky).

So, I wanted to ask you - do ceramic caps develop low resistance usually or was my cap somehow manufactured faulty or damaged?

[TerraHertz]

Are you sure it wasn't just some conductive surface contamination, of the component or PCB? Not far between the end metalizations on 0603.

That's kind of the opposite problem to one I've found several times now in cheap switchmode supplies. Some of them use a 1M resistor for a trickle start-up of the switcher on the mains side. So there's an approximately 1/4W metal film resistor with near full rectified mains voltage across it. They eventually go open circuit, so the supply never starts. Even though it would run fine if it did start. And because it's a 1M resistor, and the surrounding circuit has more leakage than that, with the resistor in circuit you can't tell it's open.

I think it must be a kind of electro-migration of the metalization on the resistor.
Replace it with two 470K R in series, and they work.

[JuKu]

So, I wanted to ask you - do ceramic caps develop low resistance usually or was my cap somehow manufactured faulty or damaged?

No. Yes. (I'd guess mechanical damage by flexing the circuit board.)

[free_electron]

Multiple possibilities. Cracking is the most likely cause. Body cracks and this causes an internal short.

[SeanB]

Ceramic chip caps are very prone to shorting, often as a result from ESD ( there is a NASA study about this) that causes a weak spot that grows whiskers, or from cracking from stress during handling, soldering or mounting of the board. They were famous for this in early encasulated ceramic caps which could burn a hole through a board when they failed.

[Pentium100]

I did not see any surface contamination.

The power supply is not a cheap one (a new one costs ~$150 IIRC) and it worked for some time before failing.