This post is only going to be relevant directly for those people working on the same kind of stuff that my group does. Still, it gives a flavor of the challenges that can pop up unexpectedly in doing experimental work.
Often we are interested in measuring the electronic conductance of some nanodevice. One approach to doing this is to apply a small AC voltage to one end of the device, and connect the other end to something called a current preamplifier (or a current-to-voltage converter, or a glorified ammeter) to measure the amount of current that flows. It's possible to build your own current preamp, but many nanodevice labs have a couple of general purpose ones lying around. A common one is the SR570, made by Stanford Research. This gadget is pretty nice - it has up to a 1 MHz bandwidth, it has built-in filter stages, it is remotely programmable, and it has various different gain settings depending on whether you want to measure microamps or picoamps of current.
Here's the problem, though. One of my students observed that his devices seemed to fail at a surprisingly high rate when using the SR570, while the failure rate was dramatically lower when using a different (though more expensive) preamp, the Keithley 428. After careful testing he found that when the SR570 changes gain ranges (there is an audible click of an internal relay when this happens, as the input stage of the amplifier is switched), spikes of > 1V (!) lasting tens of microseconds show up on the input of the amplifier (the part directly connected to the device), at least when hooked up to an oscilloscope. Our nanoscale junctions are very fragile, and these spikes irreversibly damage the devices. The Keithley, on the other hand, doesn't do this and is very quiet. Talking to SRS, this appears to be an unavoidable trait of the SR570. We're working to mitigate this problem, but it's probably good for people out there in the community using these things to know about this.