I.e. when Victoria's Secret gets hit with $10 million a year in royalties (or whatever the number was), they can say "We're not going to fight these patents in court, but if somebody else fights them and wins, so we don't have to pay that $10 million/year royalty anymore, we'll give them a one-time $3 million payment."
Victoria's Secret has an incentive to post a bounty under these terms, because having that promise available can only improve the chance that someone else will take it on. And they only have to pay out the reward when Newegg wins -- which means setting up the bounty didn't cost them money, it saved them money (assuming the reward was what made the difference between Newegg fighting and rolling over).
This way, each business can settle or fight depending on their specific circumstances, but the companies that choose to settle can still encourage others to fight.
They've all pretty much been reduced to "X on a computer" or "X over a network" with "X" being ideas that are in many cases decades old and if patented, long since those patents have expired.
I guess this is what we get with patent office incompetence and rubber-stamp-happy inspectors. The patent office a little while ago asked for "feedback" on the software patent issue. I had a good chuckle.
This kind of reasoning also applies to border skirmishes, contract disputes, and patent trolls.