- I seriously doubt that there is one, unified, conspiracy-theorist-satisfying explanation. Even if there is one, I seriously doubt that it will ever be aired in public to the satisfaction of conspiracy theorists everywhere.
- The mechanism appears to most probably include a combination of several factors, including some astroturfing, some incompetent programming and systems management, some misguided marketing and inventory-management bullshit, some inadequate training/preparation of customer-service droids and PR dorks, and the stupidity of making this into a "holiday weekend." All in all, this was a disaster of some kind waiting to happen... whether triggered with malice aforethought or otherwise.
- To the first-time customer, it doesn't matter whether there was malice aforethought involved; he/she won't get an accurate search result. This really does two things: It points out the idiocy of reflexive, black-box "sales ranks" as the primary ordering algorithm for search results, and it points out that the customer is not always right.
Most importantly, though, there's a structural cause-and-effect relationship from the positive feedback loop. Ultimately, the problem is that Amazon dominates a certain aspect of book distribution — exacerbated by the holiday-weekend nonsense — to an extent that strongly implicates antitrust issues. This is a result of the overcorrection in antitrust doctrine imposed by the Reagan administration, which chose to eliminate false-positive accusations of monopolization... no matter how many false negatives that policy engendered. The simple fact is that book distribution is an oligopoly, and Amazon exercises actual and/or potential monopoly power (such as the power to unilaterally set payment terms).
Some of the antitrust conundrum comes from difficult problems of market definition; one reaches a drastically different result for every aspect of book distribution for "category trade fiction with at least 20% of sales through chain bookstores" than for "all casebound books published". More of the problem, though, comes from the application of economic theories tested only on initial-endpoint actors in relatively clean marketplaces to midpoint and terminal-endpoint actors in the very, very messy real world. In short, we're dealing with economic theories of antitrust with similar value to the Laffer Curve: They are aggregate-behavior models (PDF) of theoretical behavior that are essentially untested in the real world. More precisely, quasieconometric antitrust models implicitly discount the endowment effect and presume actual substitution — not just theoretical substitutability — of every purchase decision based upon the characteristics of fungible commodities. And there is nothing more endowed, or less like a commodity, than individual expression.