Testing sales potential through sample orders can be difficult. Most of the sample orders I've done are for checking out the supplier's order process, customer service, and accuracy in reporting the condition of the items sold (since I mainly deal in aftermarket/overstock stuff and small dollar imports, it's important to verify that a given new supplier is reliable and truthful.) Trying to run a sales test based on x units of a given product is a difficult proposition, because so many different factors can affect sales from different weeks/months, from different venues (online and offline) etc.
If you can afford the 2,000 piece order, chances are you can afford to take a loss on it, and that a marginal loss will not cost much more than the high price and attendant costs of a market test -- ie. if 2,000 units costs you $20,000 , ask yourself if you can possibly move them for 90% of wholesale in a pinch. If so, then you spent $2,000 -- less if you had any retail sales -- and it didn't work out great, but you made a good supply connection, and possibly some good buyer connections as well (eg. other people who want to sell that item but can't swing for 2,000 MOQ, so they buy 400 from you.) That $2,000 or less that netted you a supplier relationship, some possible buyer contacts, and "physical experience" with the product is not too much more expensive than your proposed 100 piece order at $10 each plus shipping, sales/storage costs etc. and if it flies and you find you have a winner, you've skipped over the waiting period that the sample order would otherwise impose on you.
I don't generally advocate being cavalier with one's money, but I do think that sometimes people let themselves get scared by "big numbers," or worse, trapped in small numbers, because they justify numerous perceived low-risk expenditures based on small price tags without noticing that they are gradually spending themselves out of business. A low nominal cost in terms of dollars-for-merchandise must be combined with the price of the irreplaceable time, energy and focus that have to be spent moving all those test items before its true cost can be calculated.
Just my $0.02 (or at least, a $0.01 sample of same.)
Frank