[CWuestefeld]

I'm a PostgreSQL newbie [1]. From the discussion below, it seems that most of what's being addressed [2] is only really relevant in cases where there are really long-lived transactions. So a better-behaved system in which the transactions are of fleeting duration wouldn't see as much benefit because they don't suffer much from the problem.

Am I understanding that correctly?

[1] Our DBs are currently all SQL Server, but due to licensing costs, we're going to do all new stuff in pg, and over time, migrate the legacy stuff.

[2] I see there are other across-the-board benefits from e.g., smaller tuple headers; I'm talking about the main thrust.