In part I, I babbled on a bit about D&D's niche protection (as I see it) and niche protection in play.
Here, I'm going to babble on a bit about how I see the role of niche protection when designing games. I've written two official templates for GURPS Dungeon Fantasy 12: Ninja (the ninja and the assassin), and in both cases I had to give very special attention to niche protection.
All of the below assumes "all things being equal." This includes points, choices, levels in games that have them, etc. In its own niche, a very low point character might outdo a very high point character who fills another niche - but it might not. A very high level magic-user in D&D might well be better in hand-to-hand combat than a low-level fighter. The best fighter in a 100-point GURPS game might be weaker than the worst fighter in a 500-point GURPS game even if the latter isn't a combat specialist. In these cases, all things are not equal.
What I mean by niche protection in design is making sure that the putative "new" template* does a couple of things:
A) It does something better than other templates do. This is the "niche" part.
B) It doesn't do things other templates do as their main strength (their niche) better than those other templates. This is the "protection" part.
If a template can't do both A and B, it's not worth making. If it does A but not B - in other words, it out-does an existing template at that template's own main strength, you're basically replacing the old template. If it does B but not A, then it just sucks - it's not stamping on anyone's niche but it's not good enough at its own niche to actually carve it out, it's just being some redundant capability. It needs to both have its own strength and not out-do another character's strength.
In other words, you don't want a template that out-knights the knight or out-barbarians the barbarian. But you don't want one that doesn't out-something everyone at something. Whether that something is worth doing at all depends on the campaign and the players, of course.
Let's take the GURPS Dungeon Fantasy 12 assassin.
What's the assassin good at?
Well, he doesn't out-fight the fighter types. And he's not built to steal like the thief.
He's built to be the best sneak-up-and-stab-you guy in Dungeon Fantasy.
If you need someone or something snuck up on and attacked by surprise, the assassin does it very well.
He's also not so bad with poisons.
If you need traps disarmed or set, he's good at that, but not as good as the thief. He's not a bad scout, either, but urban environments not wilderness. He'll do as a fighter, too, but not as well as the dedicated fighter types. There is some of what I think of as JB's good redundancy).
So you've got a character that does something better than others, but doesn't make any other equal-point character totally redundant. You can bring an assassin instead of a thief, but he's no locksmith or fence or superior searcher, and it'll take a fair amount of points to make him one - points a dedicated thief could use to stay ahead. You'd have been better off with a thief if you needed on. A wizard could make himself invisible and use Deathtouch to kill people by sneaking up on them, but he'll need a considerable investment in points to pull it off, and he's not terribly well built to survive his sudden appearance next to his victim, and he's not going to do well against a magically resistance target or one in a NMZ. On a dungeon delve, he's generally going to need his magic for other things, and need it often - garrotes don't need quite as much magical energy to power as attack spells.
Now, a niche doesn't have to be one thing. Just something. That "thing" could be "things." You could cover a couple of areas in combination in a way that other templates do not. The titular ninja of the book above does that - fights well, like the fighter types, but with different weapons; has special powers, like the bard and martial artist - but different ones; and uses gadgets, like (but not as broadly as) the artificer, but limited to "ninja" gadgets. He doesn't do any of these "better" but rather "differently" than the others.
You might say, "so what?" to all of this. "Why do I care?" Because a poorly designed template, one without a niche or that doesn't respect and therefore protect the other niches is a covered pit trap waiting for a player to fall in. Oh, you chose the Knight? How unfortunate for you - you chose poorly. No one wants to drop into a game and find out that they can choose from A, B, and C but actually B is crud and C does everything B does but better and more. A bad game, IMO, gives you unequal choices, and worse, hidden pitfalls for people who don't know it well.
Anyway, that's the kind of approach I've used to design templates for GURPS. I'm not claiming in any way this is the only approach to use, or the official GURPS or official SJG approach, either. It's just the thought process I use - and one that seems to be echoed by the comments and questions of my editors. And it can be painfully obvious when someone didn't do this - as many poorly designed splatbook expansions can demonstrate. The new book comes out, and the no-niche-protection whatever rolls out and makes your original basic rulebook guy useless and retroactively poorly designed.