Up to 999,999,999,999,999 records? of 536 bytes each?
OK, it is really, really, going to help if you can identify the header without using the sequence-number approach, if you have a file even remotely approaching that size.
It would frankly be "nuts" to pass through some huge amount of data, just to update the sequence number for such a file.
I think you need to get some clarification. Find out the expected initial size of the file, dummy up some records, find how long it takes, take file growth into account and do some estimates for one year, two years, whatever, then go back to the requirement-giver and see if they are happy.