I think you will find that the price for the KineMAG are not out of line with their quality factor. In general you may find that you will need a drive rated about 5x (or more?) the data rate being recorded, in Flash media their record rate is uneven because of a delay while the chips bank switch, although the camera has a buffer the drive needs to be fast enough to write faster than the average rate because the delay for bank switching of the flash chips in the drive causes the buffer to fill, so, the average write speed is not telling you much from the specs, it needs to be the worst case write speed probably. KineMAG use (so I was told) high quality FLASH chips that don't slow down much with repeated read-write cycles, other drives may work when new but slow down with use and eventually not work as well, and its no savings to have the shooting cut-out in the middle of an important shot because you tried to save a few bucks, that in the end will cost you more in lost time or even an irreplaceable once in a life time shot. In the S35 model you can dump to any speed SSD or HDD in the second drive slot, and re-use the KineMAG to save some money on the higher priced KineMAG s.
I've suggested to them to add dump to USB port support in the Mini, like the way the S8p and S35 work with their dual SATA slots, so people can empty the KineMAG in the field using a USB to SATA adapter, or just dumping to a USB thumb drive, not everyone's cup of tea, but useful where you don't have a laptop along to dump the KineMAG for re-use while shooting in the field.
As for the 96fps and 100fps modes, you can look at the reviews of the camera to see what maximum record time other users are getting with the KineMAG . The maximum record time on high speed modes may not be due to the SSD speed, but rather the SATA port speed itself, or other internal bottlenecks in the data path, you can contact Kinefinity.com and ask them what maximum record time they have achieved so far, and which model of KineMAG that was achieved with to compare to the results you are getting with some other brand of SSD.
On the subject of clipped highlights, I forgot to mention that de-Bayer programs based on Adobe SDK or DCRAW etc. may clip all three color channels noticeably for shots made with low average exposure because of their internal AGC effects. Such programs may clip the top 5% of pixels (etc.), but those ARE the highlight detail that people pay more for in getting a camera with good dynamic range, so its nuts to use programs that clip of the upper pixels that are withing the ADC range on the sensor. My de-Bayer program can be setup not to clip anything, or to just clip the chroma and not the luma. You may be able to use the free program DNG_valadate.exe from Adobe , its in their DNG development zip file, to produce a gamma 1.0 conversion of DNG to TIF to check if your normal processing workflow is clipping anything in the highlights, if that does not work my de-Bayer program is also free and can also make a gamma 1.0 conversion from DNG to TIF using the engineering processing methods for checking those sorts of losses etc. In this case, programs like XNVIEW should probably not be used to check for losses of highlight detail as such programs may be clipping the upper highlights off and not give results that are representative of the upper limits of the recorded data in all three color channels.
Please keep in mind that a raw recording camera is unlike a camera that records balanced RGB data in that the results you see are NOT what the camera recorded, but are dependent on how you convert the raw recorded data INTO balanced RGB files, and also the clip limits and gamut of the monitor or projector you are using for viewing the balanced RGB files your workflow produced from the raw recordings. In many cases what is being compared is not the cameras intrinsic qualities, but the way the various workflows interpreted th raw recordings, so very different results may be obtained using the SAME camera and different workflows (and or monitors and projectors) and de-Bayer and LUT settings within the same workflow.