Hi
If I'm not mistaken, somewhere, sometimes ago Tom Lane mentioned that he's working on the Tablespace on his personal TODO list.
I'm just throwing idea here, perhaps Tom already think of it. But I send it anyway.
If the Tablespace is the same concept with Oracle then it can take advantage of clustering.
Searching for certain data with Seq Scan does not mean it searching the *whole* table but only the part of the table that resides on the certain cluster, if the key is part of the cluster key.
Best regards
Andy