I mean, you're answering your own question as far as "Why isn't this behavior being dealt with more harshly?".
It is dealt with harshly. First offense for trolling is a 14-day ban, second is a permanent ban. But, as you pointed out, there are people who troll who try to make that behavior hard to spot.
Is Riot going to catch all these people? Of course not. The anivia example I gave is an example where the person was trying to avoid the system by making sure that he didn't die a lot but was banned anyway. Likely a manual review. But no matter what sort of realistic effort Riot makes they are not going to catch 100% of trolls. No system by any game company is going to catch 100% of the people who troll, not even close to 100%. It is compounded by several reasons;
They don't have the man-power to manually review every single report. There are millions of games played each day alone.
They are under extreme pressure to not ban anyone by mistake. Banning an innocent is far worse than letting someone who's guilty skip by one time. If a guilty is missed for one instance of poor behavior they have another shot at catching them at the next instance of poor behavior, but if an innocent is mistakenly banned they likely lost the customer. So all it takes is reasonable doubt of someone to not be punished.
A automatic review is never going to be very accurate at detecting trolls. League is a complicated game and there are plenty of subtle ways that someone can troll that, across millions of games, will look indistinguishable from someone who's just playing poorly to a robot.
So what do you suggest Riot do at this point?