They should be able to update the mechanics and, when absolutely necessary, the flavour.
Over the years, problems in the lore of settings will become more apparently. Like the value of steel pieces in Dragonlance. The size of the continents of Ansalon or Khorvaire. Sometimes population numbers don't work or conflict with the tone.
(To say nothing of elements of the world that might have grown more offensive or problematic over the last few decades…)
There should be the freedom to make small necessary tweaks.
Dark Sun 4e actually has a couple of my favourite examples. Dark Sun is all about the desert and features almost every type of imaginable wasteland. But, in the 2e version they missed one. There was no barren volcanic plain. So an otherwise unremarkable stretch of land was tweaked in flavour, giving the world more variety.
Second, there is the death of the Sorcerer-King Kalak. In 2e, Kalak was always a dead man. The metaplot was always going to off him, so the campaign setting was always going to be "wrong" compared to later accessories and adventures. I think 4e made the right call of ignoring the later novels while also starting after the interesting event of Kalak's death that upsends the status quo of the setting and generates more adventure.
Because the originals are never going to be perfect. They were written for hard deadlines with firm wordcounts with limited adventures set there. Future products shouldn't be bound to lore than wasn't played and seen through the eyes of the players.
That said… I'm a canon whore. I'm always going to default to canon whenever possible, and favour additions to canon rather than changes. Problems should be fixed by making the world richer.
I agree that all races and classes do not need to be included. What options are omitted can help define a world. The "world without elves" and such.
Especially when certain races were purposely omitted back in the day. Orcs and gnomes from Dark Sun, orcs from Dragonlance, etc. Adding them back in would be a mistake.
That said, new races should be considered. The point of an update is to incorporate the new into the old. To look at the new species and say "do these fit?" Sometimes, like dragonborn and Ravenloft, the answer is "hells now, that's just wrong". Others… like tieflings and Ravenloft seem interesting and fit the tone of the setting almost better than default fantasy (if, perhaps, with a subtle appearance tweak).
You have to ask "does this species work with the setting and enrich its lore?"
For example, the idea of a wood, stone, and obsidian warforged golem sounds badass. That seems like it would work and be the kind of guard a Sorcerer-King would love to have. Warriors that they don't need to feed and that can guard their palace endlessly.
They could be a new invention that is unfamiliar, or an old one that was abandoned because the warforged lacked enough loyalty (you can't maintain obedience by withholding food from a warforged).
