Here's another chance to read this great article by Phil Rhodes on why cameras need global shutters. Cheaper, cinematic cameras come with a cost - they tend to have Rolling Shutters, which means that rapid movement can be skewed. The ability to buy cameras with global shutters at all price points can't come soon enough, according to Phil Rhodes
Shooting strobe lights at the moment is a quick route to frustration, assuming you're using one of the currently-available cameras that doesn't quite read all of the pixels from its sensor simultaneously. And that applies to the vast majority of modern cameras: Red weren't quite the first, but they were certainly the people who made the cost savings of a rolling-shutter design a part of the modern world of cameras. On the face of it, it's difficult to see this as a benefit, since skewed motion and flash banding are unattractive artefacts of cost, rather than quality-focussed engineering, although the very high resolution sensors currently in vogue would not have been available, or at least not so soon, and not with such high dynamic range, had this compromise not been made. I'm still not sure it's a very good thing, but the motivation behind the decision is clear whether you think it makes sense or not.
The situation is now fairly well understood, but briefly recap the, rolling shutter is an artefact created by the amount of time taken to read pixels from an imaging sensor which is still exposed to light. If the subject changes during this readout, the remaining rows of pixels in the image will be exposed to the changing light. As a result, objects moving horizontally may appear to be skewed somewhat diagonally by an amount proportional to the relationship between the speed of motion and the readout timing of the sensor. Some cameras, such as the Arri Alexa, use rolling shutter but have extremely brief readout time, minimising (though not entirely removing) the problem.
An old problem
Strictly speaking, rolling shutter is an old problem. The mechanical shutter of a film camera can't possibly move as fast as the progress of the electronic readout down an even reasonably fast digital sensor, but it turns out that this doesn't matter much in reality. The film shutter is in an out-of-focus plane which means that the sharp, horizontal edges of flash banding are softened to the point of almost complete invisibility. In essence, a mechanical shutter fades the image out, albeit with some areas of the image fading out rather faster than others. Strobe lighting was used, for instance, in some of the action scenes of Aliens, without objectionable artefacts. Digital cameras which use rolling-shutter sensors may also use a mechanical shutter (F65 and certain Alexas do this, as did D21), which alleviates the problem. A common test approach is to shoot a revolving drum painted black with a white vertical stripe. On rolling-shutter cameras, the stripe appears to become something of a spiral wound around the drum.
Probably the worst possible subject for a rolling-shutter camera is the xenon flash tube, as used in strobe lights and camera flashes. The discharge is extremely brief and with an extremely fast rise and fall time, from tens to a few hundred microseconds, depending on the desired light output and the way the flash tube is being driven. Even the very fastest electronic rolling shutters do not operate quickly enough to remove all possibility of the light going out halfway down the frame. Strobes based on LEDs are only slightly less of a problem as their rise time, due to the activation delay in the phosphors used in white LEDs, is very much longer; even so, it may still paint a fade of light across several dozen to a few hundred lines of the picture. Shutter-synchronised strobe lighting, such as Outsight's CreamSource LED array, can avoid the problem by reading a video signal feed from the camera and ensuring that the light never goes on or off during the shutter readout period, but that's an expensive and potentially time-consuming workaround for a newly-invented problem.
Ostensibly flicker-free lighting can also cause issues. A fluorescent tube driven at – say – 24,000 pulses per second might present no problem to a traditional 24fps camera, which would see roughly a thousand flashes per frame (some types of lighting can produce visibly segmented motion blur due to this, but it's reasonably unusual for this to be objectionable). Even if a series of frames either included or excluded one complete pulse of light due to relative timing jitter, the maximum brightness variation would be one thousandth of the light's output (and most real ballasts go many times faster than my easy-maths example). This is the situation that exists with almost all electronically-ballasted lighting and isn't complained about. But the readout time of a rolling-shutter sensor is typically many, many times faster than the frame rate, and as such, even very high frequency lighting may be read as a visible series of horizontal stripes or bands down the frame, as its flicker interacts with the readout time of the sensor. Tungsten lighting doesn't usually do this, but some ostensibly flicker-free HMIs can.