So, you have table X - it has 2M rows (say, 0.5 GB) in the first file (along with all the other tables). The 2GB limit is hit, more data is added. 0.7 GB is added to table X - these records go into a new database file - the table is split in two - you have 2 "extents" of 2GB with X split 0.5 - in extent1, 0.7 in extent2. All mixed up with other tables as well!
That was the architectural flaw to which I was referring. Nothing to do with PG
You're gonna be in for a very rude surprise after adding 1GB of data to a PG table...
