It's important to talk about distinguishable pixels rather than "4K." After the lens, the OLPF, the sensor, the RAW converter, the codec, and the necessary post, the number of pixels you can get out of a capture system is usually far less than the theoretical maximum imposed by the nominal pixel counts. Same with bit depth, sampling rate, etc. We need standardized chart/DR/RS/Noise etc. tests of every video mode of every camera but those are not to be found for reasons that continue to baffle me.
And for practical considerations, here's a nice hands-on with the FS7. Seems it's sort of at the "Mark I" level, needs firmware improvements, as does the Metabones Speedbooster to work with it. It also could use some hardware tweaks. It's a little early still for the FS7, and the Canons still have a few advantages as he mentions.
http://www.newsshooter.com/2014/11/01/sony-fs7-field-report-week-1-the-canon-c300-killer/