Recently I've been preparing to manufacture a product which exclusively uses MLCC capacitors across the board. It integrates an onboard buck converter which uses them, and MLCC are also used for local decoupling.
My prototypes have consisted of "dodgy" reflow techniques using a hot plate. Generally, 10% of the time after doing this, I find a shorted MLCC on the board, usually found because when power is applied, the cap will smoke.
However, just right now I was replacing one of these caps with a soldering iron and after I had replaced it, it was still shorted. I verified there was no other short circuit on the board (because when removed 3.3V showed a few kohms of resistance.) It seems the simple operation of soldering the cap has caused it to fail.
I've also recently repaired an LCD monitor which had a shorted MLCC on the T-con board and a few other users on a popular forum have reported this issue as being surprisingly common. Now, in this case, a monitor gets warm or hot, but no where near as hot as a soldering iron - so why could these be failing?
I am planning to offer a five year or perhaps longer warranty on these boards, but I can only do so if I am confident the board is capable of surviving normal conditions.
Caps are 0603 (100n, 10u 6.3V), 0805 (22u 6.3V) and 1206 (10u 35V). All are X5R or X7R. There are some 18pF caps for the crystal, but I've never seen those fail - I suspect they are a different technology to MLCC though.