the backlog theory is not right, what will end up happening is that [ ... ] TX fees will go up such that the TX / sec is below the 1MB limit, if we get a surge of TX demand that will only push fess higher again making low fee TX never confirm. so basically there will never be an ever growing backlog, only ever growing fees.
at one point tho the high fees will start to affect TX demand negatively ( no one wants to use bitcoin any more because its so expensive ) and at that point bitcoin growth has maxed out due to technological limitations ( or self imposed nonsensically 1MB block limit )
of course higher fees will always "affect TX demand negatively" what i mean to say it TX fee will stop simply weeding out low value TX from users, and start weeding groups of users all together.
That is the hope of the small-blockians, who look forward to the "fee market" as a way to raise the fees "naturally" to compensate for the harlving of the reward.
However, I don't expect the fees to rise significantly. After some initial mega-jams, things should settle down with the
average traffic somewhat below the capacity (say, 2.6 tx/s, as in my example). In that state, the minimum fee will usually provide confirmation within a day, and confirmation in less than 1 hour will
usually be achieved with relatively modest fees. The traffic will not increase further because of the unpredictable delays, which will occur even if everybody ends up paying 100 $/tx of fees.