It will be hard to build a reliable SSD without access to the NDA documentation from the NAND manufacturers. The datasheets you see online are only the basic information.
I don't know what you mean, but I use (and SW-integrate) NAND chips all the time and complete datasheets and application notes are available for most known brands.
Try looking for the datasheets on any modern high-capacity 3D NAND of the sort used in SSDs. What you've linked to is an ancient part with 1/32 the capacity of the smallest NAND that can be found in any recent SSD.
Yes, the linked one is the first PDF I've found. I know the commercial NANDs found in newest SSD may not be available, because it's probably that the SSD is using a custom chip design. Perhaps the chip is listed somewhere in a website, but unavailable for purchase (for mere mortals like us or those doing an open source SSD). In any case, those chips won't be available for an open source SSD.
Soon I should put hands on a design with a 2Tb Micron NAND from 2014 (MT29F256G08something). The PDF is behind a simple user registration.
SLC NAND is pretty reliable. Also the older gen MLC NAND is not too bad. But around 60nm down things get progressively worse. And moving from SLC to MLC to TLC to 3D it gets even more.
Page 10 shows the 3 generations of OpenSSD they've been developing. They're on the 3rd gen now (marked 2016), which is FPGA based, says it support a subset of NVMe, and mentions a 2TB max capacity.
Might be interesting to actually do stuff with. :)
They are using Xilinx Zenq 7000 series which also have ARM core and memory controller along the FPGA core, think all major electronic parts sellers are selling them and not expensive part.
Interesting project non the less
[0] https://en.wikipedia.org/wiki/Indilinx