In the old model, the way to get your game onto Steam was to put it through Greenlight, where if it gathered enough support it could get promoted to the main storefront. You had to pay a fee to Valve to submit games to Greenlight, but it was a one-time fee ($100, IIRC), after which you could submit as many games to Greenlight as you wanted.
In the new model, Greenlight goes away; you just submit your game directly, and after you pay a publication fee, the amount of which is still up in the air (anywhere from $100-$5,000), it goes into the storefront. But, and this is the wrinkle, now you pay the publication fee for each game you submit rather than just paying it once. So if you submit ten games for publication, you pay the publication fee ten times.
So why the change? My guess is that Valve has decided that the essential premise of Greenlight -- that community curation would surface the gems within the vast pile of dross that gets submitted -- has been comprehensively broken by tactics like voting rings and giving out free keys in exchange for votes. So this represents them abandoning that idea and replacing it with a more blunt filter: a per-title publication fee. In the old model it cost the same to submit a thousand games as it did to submit one, so developers just threw lots of things against the wall and prayed one would stick. Now there will be a financial incentive for developers to put more wood behind fewer arrows, as the saying goes.
The big question is whether this will improve the overall quality of games on Steam, or reduce it further (if that's even possible). I'm honestly not sure. A lot will depend on where precisely they set that publication fee -- too high will drive off cash-starved indie devs, too low will mean the crap merchants will just swallow it as the cost of doing business.
I'm dismayed there is scarcely any mention of the quality of works entering Steam. It's patently obvious that quality is the major issue now, not the usability of the platform or developer features.
It seems pretty clear to me that their plans are to rely on the market to moderate this stuff.
My second concern is that Greenlight and shovelware will turn into a malware channel. Some of the anti-cheat utilities have deep access.
This is a case where the developers had a vision (and still do), but early access to a mode that was initially just for testing allowed them to discover their test mode (aka "sandbox") was actually a fine game in and of itself.
Early access, I think, allows developers with perfection obsession to go ahead and ship something. Instead of developing it forever, hoping to achieve their goal, and never shipping.
It's true that there is a lot of garbage in early access, but I don't think it's a worse ratio than most indie games.
Bad games don't make much money before it's discovered and Steam has a 14 day-ish refund policy. It might force them to automate it, like Google Play does for 2 hours after purchases.