Unfortunately, one of the truths I learned early on as a software engineer is that every time you 'fix' something, there's a non-trivial probability that you'll break something else, possibly quite badly. To try to prevent such broken code reaching the market, every new version needs to go through large amounts of regression testing.
Do you think WotC are really doing that sort of testing each time they throw out their latest batch of changes?
Well, no. But I don't think the game design is nearly as interconnected as your average piece of code. While fixing a power can lead to unintended consequences, or a piece of errata could go too far in the opposite direction - making a weak option overpowered, etc - I don't think that happens nearly that often.
Do they occasionally need to errata the errata? Sure. A small part of it. Just like a new book like Martial Power might offer 1000 new options, of which 5% (50) need errata... those 50 pieces of errata might have a fraction that need fixing. Given the shorter turnaround, maybe at twice the usual rate of errors, so let's say 10% of those updates themselves need errata - or 5 items. And once those are fixed - which is probably done with extra care - we should now have most imbalances taken care of.
The notion that each batch of fixes makes the game better is faulty. The game as it currently stands is almost certainly no better balanced than it was when first published; it's just that the unbalanced areas are in different places.
I disagree entirely. Comparing the game as it currently stands to at first launch ignores the fact that in addition to the errata, you've also many books filled with new content.
If you instead compared the PHB at launch to the PHB with current errata? Then yes, it absolutely is a far better balanced book.