In the software world, scaling comprises two parts, the technical aspect of whether the application will scale as users grow, and the marketing aspect. The techical part is that it wont require an exponential increase in server / computing capability, and ideally, such costs per user would drop as more users are added. The marketing part, is the marketing overhead per user drops as the application grows.. always a tricky part with any type of business, but with open source, perhaps even more critical.
In the hardware world, economy of scale also comprises two distinct parts. First, raw materials/components prices drop as the line item purchases become larger. Ie, if I want to buy one rca jack, its $2 at Radio Shack, it I want 1,000,000 of them, I can even get them customized for a fraction of that cost. Its even more dramatic with enclosures, ie getting 5 thermoformed enclosures might end up costing $500 each, where as getting 500,000 injection molded enclosures might drop the cost down to $0.50 or less.
Then there is manufacturing overhead… everyone would like to build small quantites economically, but when it comes to electronics, often times the setup costs are tens if not hundreds of times the individual piece part costs. Ie, it might take 20 seconds to populate a large dense circuit board, but it takes 4-8 hours to program, load, and test the assembly equipment the first time to make it so. A similiar deal exists in test engineering… ie, a board level test fixture costs $2500, and whether you run 10 or 100,000 units through it, (assuming one already has a lab view style master test console) the fixed costs remain the same. Lastly there is the knowledge base of the line technicians… a 10 piece run does not develop a knowledge base to allow fast rework/repair, or troubleshooting, where as a 10K run pretty much means the line techs are fully up to speed and ready to roll.
All of these factors taken together, make small production runs of open source hardware problematic. Granted, if the margin is there either by uniqueness or customization oppurtunities, its much less of a problem, but for low margin products, its a real challenge.
Some of the ways to mitigate this, are to choose parts which keep the bom cost at a minimum to start with. Ie, avoid $25 highly specific parts, even though the prices drops like a rock with volume. Another solution would include test fixture designs with the design, such that test engineering overhead is minimized. And of course, using production notes to get line techs up to speed, well before they have run 10,000 plus parts. It may be that the use of a collaborative wiki where all manufacturers can chime in with ideas, problems, and fixes may also be of great help in keeping the economy of scale manageable for low volume production runs.