^^ what he said. Having said that after you've loaded and setup a few boards like that you should soon realise on small batches how much of the cost if you were subbing out was going on setting up.
Stripping down and starting over would require careful planning, you would need to run through your parts in height order and kit them to yourself split into the mini jobs, way to save a rummage fest for each loadout. Every extra line on a BOM is another component to buy, another component for RS/Farnell/Mouser to pick incorrectly,another part to load, another part to unload, another part to store and label all of those things take time and if you're working in small qtys and strips of tape come with a massive added risk of getting mixed up. Parts that come in larger tape sizes also use up proportionally(usually) more "slots" on your placement machine, 48 feeder slots really is not very many. As Mike points out the neoden feaders look fiddly (essentially fit once never move, I wouldn't want to be moving those terminals that often), so one of the rival machines that uses a more generic removable style might be better.
Consolidating the BOM: the first thing I would look at is pointless "cost savings", in low volumes the component cost of passives is basically irrelevant so if you have a bunch of special values look at why.
Examples of this can (IME) include having the same value cap in multiple voltages or the same resistance but some are 0805 to handle a power requirement. Next you can look at weird resistor/cap values - maybe you can make some of them from more standard ones already in the design. I would go back to the basic design itself last, check the specs/tolerances, arbitrary pullup/bypass choices etc.
More extreme measures could include splitting out bits of the board as modules, handy if you frequently reuse bits across multiple designs and can sometimes make it easier to test/fault find.