Note: this is gonna be one of those long technical posts. I started writing several days ago, adding a bit here and there as I had time. So it grew and grew, and I don't see any parts that can really be thrown out. So read only if you want to see inside of a certain sort of operation. It's pretty much limited to portrait films (aka Portra, albeit an earlier version) under electronic flash.
It's really quite amazing. [macfred's photolab link] Though I would like to see the results with darkroom prints and with other scanners. Sometimes I wonder if the magic of overexposure is in the film or...
I've been through that routine upwards of a dozen times over the years, mostly with the 125 or 160 speed Kodak professional portrait films. Printed optically onto the appropriate pro paper. This was done within the main processing lab of a large studio chain outfit. We were especially finicky because the film was all ours, and with the amount we used it couldn't be bought "off the shelf." There were contractual obligations well into the future. So if you committed to something that didn't work out, you couldn't just change your mind.
Consequently we did some pretty extensive testing with a variety of complexions, from very pale to very dark, as well as different hair colors. And we did this over a wide variety of exposures, not because we intended to use this range, but because of the possibility of exposure screw-ups in the studios. With a large number of studios it was a regular routine for this to happen periodically.
Basically we would test with about 4 or 5 different models - different ccmplexions and hair color - with a variety of colored clothing. Some of the clothing we supplied (x-tra large shirts, worn over clothes as smocks), as well as a handful of colored fabric samples. These gave us the ability to duplicate some clothing colors over the years. We also included measurable color reference targets in the scene.
Now, if you're going to do a test where you're finicky about exposure then you have to be specific about what "correct" exposure is. All of the mainstream pro portrait film makers supplied such info in the form of aim densities for gray cards and flesh tones. (Additionally, the manufacturers used to supply "printer setup negative sets," known on the internet as "Shirleys;" the middle negative was defined as "normal" exposure, so one could use this as an analyzer (such as Kodak PVAC) or printing data reference. When we used known-calibrated Minolta incident flash meters under mostly frontal (standard-style) studio lighting the film exposures came out right on the money. I guess it goes without saying that the processing has to also be right on the money, otherwise this can put the film densities off; we always had a "process control strip" run with our film tests.
For printing, we'd use a negative close to normal, color balance to "ideal," which was a committee decision by 4 or 5 well-qualified people viewing in a color-correction booth. After this, each negative being evaluated is hand balanced to closely match the "normal" exposure neg. (As a portrait specialist we did our matching on the flesh highlights; everything else falls where it will.) Finally, everything is evaluated in the color booth - are we having color crosses in any of the skin tones, where do the exposure extremes begin to suffer, how are the reference colors coming out? And other similar things. Our goal was to shoot huge amounts of portrait sittings in our studios, then to print various package assortments in what is basically a production line sort of operation. Most of the time, I'd say that about 2/3 of our production was within +/- 2 CC units on the color, although 5 CC was our official standard. One might wonder, why the finicky testing with such a relatively "loose" standard? Well, if you try to control individual variables tighter than you need, you tend to end up with an overall higher success rate in your final product - this is, after all, a lot like a factory for pictures.
(More than) A few words about "color shifts" with changing film exposure. First, one has to consider what this really means. If you are using a manual color head, or even CC filters, and you balance on one specific negative, what happens when the film exposure changes? Everyone understands that the printing exposure has to change, but what about the color filters? This would constitute a color shift, right? And I guess it would. But... major production labs never worry about this. You might wonder why not? It's because it is automatically compensated for in every major type of printing machine. This is called "printer slope," and is (or was, in the days of optical printing) set up via the use of the printer setup negs (aka "Shirleys"). The printing machines and analyzers were set up to make matching "best" prints from an exposure series of negatives. Whatever filtration differences existed between the negatives, the printing system is essentially told, "from now on these color offsets are to be seen as the same print data as the normal negative." The printing systems then works out an internal "correction algorithm," aka slope, and henceforth no special correction has to be put in for exposure variations. In other words, the commercial machine printer never sees a color shift due to film exposure. But someone working with a manual color head likely will. Note that a major reason for slope control is considered to be due to the change in exposure time for the paper, so it probably helps to keep a constant exposure time. (I don't have much experience with this, but I imagine that Drew W knows all about it.) My point here is that anyone who worked with well set-up printers under computer control would likely never "see" such a color shift with film exposure, even though it may exist. And I am personally ignorant of the extent of such shifts since I nearly always worked with slope-controlled machines.
Finally, what was the exposure latitude on these professional films? Based on the testing I described, and only for the specific films I dealt with, the last being Portra 160 NC (replaced years ago by just Portra 160), we could run from about 1 stop under to about 3 or 4 stops over, and the resulting prints would look virtually identical. In our color booths, we could lay our actual test fabrics next to the print, and most people would say they're virtually the same color. (I'm sure that one could find fabrics that would fail this test - just saying that our selections, including strong red, green, and blue colors, were nearly a match -at least close enough that very few people would squawk about it. ) Now, if you were to leave the color booth, all bets are off. The dyes in the print are spectrally different than the original, which means that they can't match under every light. The color booth has a full spectral makeup, same as light from the sun, which is necessary to see color deficiencies in the print. (If you are trying to evaluate color prints under an "unknown" light, GOOD LUCK! At least compare under natural light for a reality check.)
If we exceeded a stop under (maybe the limit was a bit lower; I'm going from memory) the first problem to show up would be "graininess" in the darkest parts of the scene. We specifically included black fabrics in our tests, so we could see this sort of thing. If you note in the link by macfred (post #7, I believe), Canadian Film Lab's tests (or is it UK Film Lab?) you can see that the test subject is light-complexioned, not too-dark hair, and light-colored clothing. So this test subject is not gonna stress the dark-handling capability of the film; they're gonna find that the can underexpose more than if they had dressed in black, etc. You might note that in their -2 stop examples the darker tree trunk in the background has turned a different color - it has become a bit "reddish" compared with the other exposures. So it seems to be an underexposure effect. Now, where I come from, this would be outside the "usable" latitude of the film, although most casual shooters would probably say it's fine.
Another significant note is, if the subject is dark-complexioned you may want to reduce the printing exposure to lighten them up; this means it is more "difficult" to get dark-enough blacks in the print. You'll run into problems earlier with the shadow portion of the film response. More bluntly, you will have less underexposure leeway - perhaps a half stop, even a full stop or more might be lost. So in these cases your underexposure leeway might be only a half stop or less.
Anyway, this just shows some of the things that can be involved in evaluating a portrait film under ideally-balanced (electronic flash) lighting, and why different people might claim different exposure latitude limits.
I don't have any critical experience like this with respect to using these films under different light sources nor with non-portrait films, so I'll defer to those who do so.