What are the reasons to cross-process (developing C-41 film in E6 chemistry or vice versa) films? How repeatable are the results with the same film? In Flickr they all look different and you never know if this is the result of chemical or additional digital processing so realistic examples are also welcome.
Using the same film, cross-processing has perfectly repeatable results, but it helps to be specific: the negatives will turn out the same, positives (virtually always scans) may not, and it's the positives you see on Flickr and elsewhere.
The reasons for this are pretty simple: both the film emulsions and processing chemistry (both E-6 and C-41) are highly standardised. If either one changed unexpectedly to any degree, you'd get unpredictable results regardless of whether you were cross-processing.
Scans (especially of E-6 in C-41, but the reverse as well to some degree) admit much more variation - if done at the lab, it depends on the machine, it's presets, how much effort the operator takes, and whether they're familiar with cross-processed negatives. When people scan at home, they might do so with different goals in mind - some may want the extreme colour casts, some may prefer to correct the image to almost "normal", and anywhere in between. The results are like you see: the same film (even the same image) can have wildly varying results depending on its treatment.
I should add that this is true to some degree of any negative film – there is always much more room for variation in the colour balance and treatment when converting to positive than when shooting direct positive film.
Why? Good question.
I don't think there's a general answer to "why" - often it's just a matter of enjoying the look (like with any other film/technique), following a fad (they've happened a few times), or even not having easy access to E-6 processing. If there is a general reason "why," then I don't think it's significantly different to any other film. For example, original Agfa Precisa1 yielded a high-contrast image with exaggerated blues, but otherwise mostly neutral colour shifts, and an almost-clear base. As a result, it was reputedly used for commercial shots of blue jeans.
The above makes "realistic" examples difficult, but here's some mine (otherwise it's impossible to comment on processing/lighting) to display the kind of range you can get, shot on Precisa:
(from left to right)
Going back to the question of repeatability, even with such a broad range of outcomes, in the same situation I'd have every confidence of getting the same result.
1 Current new/in-date Agfa Precisa is not manufactured by Agfa, but is a different emulsion packaged under the brand. It's reported to have a greenish cast when cross-processed (I haven't tried it myself).
The results are actually pretty repeatable if (and that is important for ALL darkroom work) you keep all variables the same. You will have to use the same chemicals, the same development times and the same temperatures. Even the same drums/tanks. If you vary at least one of the variables the result get less predictable. This applies to standard color and black and wite processing too. Just not quite to the same extend because the negs and chemicals have been designed and tested for the standard processes. Predicability increases again when you get some experience playing with the variables.
If you do not develop yourself you will have to trust the lab of your choice and hope that they will keep their machines always well calibrated and stay with the same chemical brands.
But to be honest, for me the exciting thing about cross processing and the main motivation behind using it is unpredictablilty.
The reason is to achieve this special "cross-processed" look with colder shadows, warmer highlights and more color contrast.
In film photography results are repeatable only for the given exact combination of film, chemicals and the developing process details: length, temperature, etc.
In digital photography results are absolutely repeatable.
However nowadays people call almost any color distortion "a cross process", which renders the term too general to mean something.