Single-threading servers and select()

[duck <duck@...>]

Recently, it was suggested that select()-driven TCP servers could never 
scale to many 1000s of connections because of the linear cost of searching 
the select array each time a non-zero number of connections had some data
handy.

In the real world, is this really an intractable problem? With very many 
connections, surely each select() will identify many connections with data
available, thus amortising the cost of scanning the select array over many 
connections?