I believe with C.S. Lewis that ultimately it will be seen that our
whole life was going towards one end or the other, and those who fall
away were -- when seen from a sufficiently long-term perspective --
never part of the saved.  However from our current perspective, "once
saved always saved" seems like nonsense.  People do fall away.  The
NT certainly admits this, in passages such as Mat 24:10ff (which
talks about falling away due to persecution), and Mk 4 (the parable
of the sower, which shows a variety of different causes).  

Thus maintaining "once saved always saved" requires one of two
approaches: (1) that those who fall away were not saved in the first
place.  They only appeared to be.  In some sense I agree with this.
From God's perspective it is probably true.  But OSAS is normally
intended as assurance, and I don't think from our perspective it's
possible to tell the difference between "real" salvation and the
temporary kind.  I don't think those who fall away were necessarily
insincere in their original commitment in any way that we or even
they could have detected.  The other possible answer is (2) that when
we have accepted Christ at one time, we continue to be saved, even if
we lose our faith and become indistinguishable from someone who is
not saved.  I don't see any completely definitive answer to this.
(In general I don't think we are ever given an exact algorithm for
judgement.)  But I think the indications are against it.  There are a
number of statements by Jesus that suggest people are judged by their
final state.  E.g. Lk 9:52 (anyone who puts their hand to the plow
and looks back is not fit for the Kingdom), Lk 13 (parable of the fig
tree -- when the tree stops bearing fruit it is no longer good for
anything), and Mt 21:28ff (the son who said yes but didn't follow
through with it).

It is of course possible to combine these two approaches, and say
that some people who fall away were never saved in the first place,
whereas others are still saved despite their apparent defection.
However most writers seem to emphasize one or the other approach.

My primary concern about OSAS is not so much from a doctrinal
viewpoint -- as I've said above, from a certain perspective I agree
with it, but from the point of view of how it is used.  In many
cases, it's used to reassure people of their salvation.  The whole
business of trying to judge whether we are saved seems to me a
dangerous one, which quickly results in paradoxes of various sorts.
In fact I don't think Christians should live in fear of being damned.
I think it's possible to be confident that you are saved.  But most
of the attempts to describe ways of proving it seem to result in
serious trouble.

The problem with trying to establish that you have been saved is that
this involves finding tests of "savedness".  Inevitably such tests
focus on us.  But any grounds for confidence in salvation cannot be
in us -- they have to be in God.  You can see this problem occuring
in the history of the Reformation and the period immediately
following.  One of Luther's early motivations was precisely this: he
could not convince himself that he had repented sufficiently, and
thus was not sure of his salvation.  The insight that led to his
change was that his confidence must be in God, not in anything he had
done.  However this insight is remarkably hard to codify in terms of
precise theology.  It eventually led to predestination.  But that has
its own confidence problem.  If the only ground for salvation is the
fact that we are chosen by God, the problem becomes knowing whether
we are chosen.  And in fact in "scholastic" Calvinism one of the
primary issues was trying to come up with tests for knowing whether
you were chosen or not.

In my opinion the basic problem is the desire for tests to
demonstrate that one is saved.  The essence of Christianity is to
live a life centered in God.  In the context of this relationship
with God, we may come to trust in him that we are loved.  However as
soon as we try to develop a test for savedness, we end up taking our
eye off of God and focusing it on ourselves.  At this point, no
possible result is right.  It's the equivalent of jealousy in human
relationships: when one is not satisfied with the trust that develops
as part of the relatinship, but requires tests of loyalty, one is
defeated from the very start.  I think the same is true of OSAS.  In
the context of a relationship with God, we may be confident of our
final salvation.  God has the power and desire to preserve us.  But
as soon as we try to turn this into general doctrine and specific
tests, we run into problems.