Agents that did that sort of thing have been getting retrained and even disciplined by Amazon when they get caught. Amazon also updated and clarified their feedback removal section to emphasize the only reasons they are generally allowed to remove feedback. It has become increasingly difficult to get feedback removed unless it very clearly falls under one of the reasons for removal.
This guy is not just lucky in getting the negative feedback removed by bots, he has found a way to manipulate the bots into removing feedback that should not have been removed. The automated removal bots are programmed to decline or not answer requests and send them to a human for review when the seller exceeds the rate of removal requests that is deemed acceptable by the algorithm.
It likely that he has determined what will trigger the bots to pass on the requests and uses the multitude of fake positive feedbacks to keep the % of feedback that he requests removal on below that threshold. The fake positive feedback doesn’t need to be very convincing because the seller only really cares about fooling the bots, not humans. Most buyers will just order based on the percent of positive feedback they see next to the seller’s name (if they look at feedback scores at all) and don’t bother reading positive feedbacks to see if they are real.
It is also possible that he has an accomplice that works at Amazon removing feedback for him, but that is very unlikely as their performance is monitored and that employee would likely have been caught for approving too many removal requests.