Plausible deniability + army of lawyers + crapshoot of a jury trial = risk averse prosecutors. Lower-level employees who play a more direct role do often get pinched. For example, in the Libor scandal.[1]
Prosecutors don't get dinged for being risk averse. And the same is true, to a lesser extent, for regulators. But attorneys who specialize in class action civil suits can't afford to be risk averse; they make payday by being aggressive and hunting out malfeasance, which is why business hates class actions. Corporate America has successfully systematically demonized "trial lawyers" the same way they demonized labor unions. Even the most anti-corporatist activists tend to assume trial lawyer lobbyists and labor unions are more corrupt than corporations, which is dubious to say the least. (As Jimmy Hoffa said circa 1959, "Twenty years ago the employers had all the hoodlums working for them as strikebreakers. Now we’ve got a few, and everybody's screaming." That doesn't excuse lawlessness, but it should effect how we allocate limited enforcement resources.)
IMO, if we want to restore balance we need to make it easier for victims and shareholders to pursue civil remedies independently, and strip executives and board members of some of their financial protections. This way we realign financial incentives in a manner that, hopefully, maximizes disincentives for cheating (i.e. executives have to watch their back for trial lawyer "sharks"), without having to literally make a federal case out of everything. It will be costly, but corporate malfeasance is already costly, it's just that those costs are hidden--at least quantitatively hidden; qualitatively we all see it. Importantly, these costs won't need to be borne by tax payers[2] (i.e. growing ever-despised federal bureaucracy), which isn't sustainable politically. That's another reason why the SEC, FTC, and others don't pursue malfeasance more aggressively--their staffs are tiny and ill-equipped to perform investigations. It's why the SEC heavily relies on automated, heuristic algorithms for detecting financial fraud; and why reporting requirements are so invasive. We're relying on computers for lack of human resources, but computers are only good at detecting the simple schemes, the reporting costs are regressive, and prosecutions favor small fry--the type of people that taxpayers can more readily identify with, which makes prosecutions seem unfair and predatory.
Corporate America pushes back so heavily on this stuff because for 40 years large corporations have slowly turned into accounting and financing firms, with manufacturing and services as side businesses. They basically specialize in aggressive accounting strategies, which makes them feel vulnerable. And because that's just how business is done these days, legislators and think tanks have internalized the notion that permitting more corporate and personal liability for harmful accounting and financing strategies will seriously harm the entire economy, at least as measured by GDP growth. And, to be honest, for a brief period it very well might, until corporate America shifts away from accounting gimmicks and more aggressively pursues other profit opportunities.
[1] As an engineer I pitch myself to managers as someone who "gets things done". I'm an ask for forgiveness, not ask for permission kind of worker. And I assume many on HN are the same way, both in IT and other industries. So I don't doubt that executives often lack deep insight into the technicalities of various schemes. And in criminal law especially, personal knowledge of the technicalities is often a crucial element of liability.
[2] Hopefully most costs will be principally borne by the wealthy. But to the extent they're borne by consumers through higher prices or less choice, they'd be largely hidden, which is a good thing, because consumers and taxpayers aren't very good at judging long-term tradeoffs. If we can agree democratically that corporate malfeasance needs to be aggressively addressed, then let's just use the best strategy--the one corporate interests have the least ability to capture and disassemble.