1... Define your own cluster, consisting of a U32 (or a DBL) for the TimeStamp, and an array of SGL. Create the file as a datalog file, write the cluster to it, and close the file. No need for the high-level stuff. When you want to read it, open the file as a datalog file, read the cluster, close it.
2... Get the current date/time as a DBL. Use the QUOTIENT / REMAINDER function to divide it by 86400.0 (Seconds per day). Convert the two numbers to SGLs, and store them in the SGL array as the 0 and 1 items (or last and next-to-last, whatever). When reading, extract those two SGLs, multiply the quotient by 86400.0 (DBL), and add the remainder. There's your original number.
There's nothing magical about 86400, it's just an easy-to-remember (for me) number
. You could use any other number in the same vicinity (like 100,000) if you'd rather - just use the same number on both ends.