First, recognize that Level Adjustment is crippling
You describe the shade’s LA +2 as “too low” before you noticed the light-conditional nature of it. That may well be true, compared to what you’re used to seeing from templates, but in reality there is exactly one LA +2 option in the game that is any good (saint, from Book of Exalted Deeds, which is in fact overpowered). There are no workable LA +3 options, and fairly few LA +1 options. Fiendish, to say nothing of half-fiend, does not find its way onto that list.
The reason for this is because LA means your hp, skills, feats, and saves are all far too weak for your supposed level. Whatever special abilities you get don’t matter very much if you’re dead or worse because you failed a save and have far too few hit points to survive.
Worse, these problems make LA extremely difficult to DM for—the character is, in all likelihood, too good at some things and not remotely good enough at others. If you play to their strengths, they outshine everyone. If you don’t, they’re probably dead.
In short, the entire level adjustment, effective character level system never worked. We can see that Wizards of the Coast themselves recognized this, as later on they printed loads and loads of lower-LA and zero-LA options for these kinds of things—stuff that would be redundant if LA did its job, but since it didn’t, viable options didn’t exist. (None of those ideas really worked out very well either, but I digress.)
Second, making things conditional can be very dubious for balance
The problem is actually kind of similar to the problem of LA itself—you get a character who is skewed, too good in some circumstances and not good enough in others. Ultimately, this is really a question of quantity, not quality—all characters have things they’re good at and things they’re bad at. The problem here is that when we’re talking about LA, we’re talking about things that are at least nominally worth levels, and a level in D&D 3.5e is unbelievably massive. Some character classes can easily be said to more-than-double in power every two levels (when they get new spell levels), so being even a level behind might make you half the character you should be. If something is going to actually be worth that, it has to be similarly massive.
And turning that kind of power on and off is problematic because, if the condition is really representing a substantial discount, they are going to be a lot more powerful in some situations, and a lot less powerful in others. It’s just too large a swing.
But, third, fiendish and half-fiend are kind of garbage
The points I made first, about LA in general, would be, and are, true even when you have powerful abilities found behind LA. And the second point revolves around the idea that these conditional abilities, nominally worth a level or two, are actually worth a level or two—and so turning them on and off results in a huge swing.
In reality, fiendish isn’t worth a level by any stretch of the imagination, and half-fiend is even less worth four (a preposterous number).
Darkvision is nice enough, but it’s hardly amazing—light is easily obtained if you need it, and in any event darkvision is pretty common and the character might well already have it.
A 1/day smite without any attack bonus is barely worth remembering you have.
DR \$x\$/magic is pointless, since every real threat will ignore it anyway.
Spell resistance is at least as bad for you as it is bad for your enemies, because buffs and healing also need to overcome it—it very well might be best for someone to just leave it down all the time.
Really, for fiendish, only the resistance to cold and fire is any good, but there are far cheaper ways to get it (in fact, a barbarian could take Blazing Berserker from Sandstorm and Frozen Berserker from Frostburn and be outright immune to both cold and fire during a rage—two feats is a hefty price, but still better than a whole level, and in any event immunity is a whole lot better than resistance 10).
- And that’s if you rule that HD from class levels count towards those tables (it’s unclear, but a lot of DMs rule that such things refer only to racial hit dice, and in some cases—definitely not fiendish—allowing it to include HD from class levels could be overpowered).
(Actually, it should be noted, the real point of the fiendish template is for summon monster, where the big advantage of the template is giving the creature a high enough Intelligence score to speak with and give commands to.)
(Half-fiend only) The bite and claw attacks are pretty good, but easily obtained far more cheaply. Also tends to rely on being built around; a typical barbarian who is focusing on his weapon tends to not have a lot of use for a bite, and the claws are actively ignored since those hands need to hold the weapon.
(Half-fiend only) Natural armor is an extremely low-value defense, just about the lowest-value defense in the game. Too many things ignore it, the things that don’t “just” deal damage (not great, but you have an hp buffer to take it, while non-damage effects can be far, far worse).
(Half-fiend only) Immunity to poison, resistance to acid, cold, electricity, and fire, 10 each. Poisons are a rather low-threat issue, so that immunity isn’t so great, but it’s still nice. The resistances are nice too. Acid and electricity are not trivially available for a couple of feats, either. Decent. Not remotely worth it, but decent.
(Half-fiend only) Spell-like abilities—a bigger question of whether or not you will allow HD from class levels to count for these kinds of things, since this is a bigger deal than resistance or DR. But a moot point if the barbarian only gets the template abilities during a rage, since then they cannot use these as raging blocks them. They could use the rage mage prestige class to overcome that, but they’d have to somehow qualify, and then actually spend levels on it—and it’s not a great class. If they do that, they’ve earned it, I think. Alternatively, you could rule that since these are part of the rage-conditional template, the barbarian can use them during a rage.
If you do that, though, note that blasphemy at 11th level—when clerics don’t get it until 13th—is overpowered, because blasphemy is a broken spell. Most of the list is pretty meh, but that one’s incredible. Only 1/day, but that’s enough. Still, harder to pump caster level on a spell-like ability—no prayer bead of karma, for instance—, which overcomes the biggest problem with a cleric’s blasphemy.
(Half-fiend only) Flight is amazing. Eventually, flight is crucially important for every character. And barbarians do not have any native access to flight. But for an evil character, you could just get the feathered wings fiendish graft from Fiend Folio for 10,000 gp, an affordable figure by mid levels. A level is worth far, far more than 10,000 gp, to say nothing of four.
So are these benefits problematic to have conditional on a rage? The fiendish ones definitely are not. I’m not even sure I’d say they’re worth a feat. Probably they are, but I can think of a lot of feats I’d rather take. They definitely aren’t worth an LA +1, whether conditional or otherwise. With the full benefits of half-fiend, those might be worth an LA +1—flight is important, you could build around the claws and bite and that could be something, and then if you get the spell-like abilities, blasphemy is overpowered and allows you to devastate one fight a day (so long as you aren’t fighting evil creatures, which for a fiendish character might or might not be a valid assumption). LA +4, obviously not (nothing is worth LA +4).
Finally, consider an LA +0 option
The best way to handle LA options, in my experience, is figure out other appropriate costs or drawbacks to them, to replace the LA. Then you avoid all the problems that LA has.
Possibly this could be a feat
You could offer this as a feat—a rage-conditional half-fiend template is rather good for a feat, but if you tie the flight to barbarian class level rather than HD, you easily overcome that because it forces them to keep taking barbarian class levels after they have largely become pointless.
Grafts provide a good precedent for these kinds of effects
You could use the graft approach instead, charging them money for the process. You can do this in terms of WBL, too, where you don’t actually force their character to turn over gold coins in-character, but you track the value of these effects as part of their wealth and adjust loot piles accordingly, to try to maintain them at an appropriate wealth for their level, including these benefits. Fiend folio’s grafts provide a good basis for the pricing:
the feathered wings say that the flight is worth 10,000 gp,
the fiendish jaw suggests a bite is worth 2,000 gp,
the pair of claws is probably worth about the same seeing as they replace your weapon most of the time (the clawed arm graft has far greater powers, since it can attack independently, so its 50,000 gp isn’t relevant here),
there’s no fiendish graft for resistance, oddly enough, but Races of the Dragon has a natural armor +1, resistance 5 graft for 9,000 gp. Natural armor +1 is worth 2,000 gp in magic item form, but since this is a non-magical graft, taking no item slot and so on, we can easily value it as much more than that. Resistance 10 is more than 5, obviously, so that’s probably pretty expensive—one could argue for as much as 25,000 gp each, but that’s ridiculous. Still, four resistance 10’s is going to run a lot—it’s hard to argue for a price that’s really what they’re worth, since existing options for it are overpriced. I’d probably put this at 20,000 gp, total, but your player might want to downgrade or eliminate this benefit rather than pay for it.
I think it’s probably best to just leave out the DR and SR, since they’re kind of pointless but you don’t really want to give them out free. Darkvision, on the other hand, I’d be fine giving out free here as part of the package (certainly wouldn’t charge the outrageous 24,000 gp of the darksight eyes illithid graft from Lords of Madness), and likewise the 1/day smite good with no attack bonus. I’d leave the spell-like abilities out, since they were dubious to begin with.
Total, that’s around 40,000 gp, or 20,000 without the energy resistances, for permanent effects. If we tie it to rage—an effect measured in rounds and usable a few times a day—we might look to the custom magic item pricing guidelines for assistance. A continuous magic item of a rounds-long effect has a factor of 8,000 gp, while a 3/day effect—splitting the difference on how many times a barbarian gets to rage—has a factor of ⅗×2,000 gp, or 1,200. That means that tying this effect to rage should make it cost 15% of what the full price would be—or 6,000 gp. That seems much too low—these are guidelines after all—but suggests that charging 10,000 gp could be quite reasonable.
Note that Fiend Folio also has rules for the drawbacks of grafts and how the fiendish influence of grafts can affect a character.
An LA +0 template
This is probably not going to work for half-fiend—those resistances and flight are probably too good to provide appropriate drawbacks—but you could easily apply some drawbacks to fiendish to make it a roughly neutral option. Penalties to social interactions (not Charisma itself, but Bluff and Diplomacy for example), and maybe the character takes extra damage from good-aligned attacks and weapons? If everyone attacking the character with a good-aligned attack dealt an extra +1d6 damage, that would go a long way to mitigating the minimal benefits of the fiendish template. Tie the whole thing to rage, and I think that’s a quite-reasonable LA +0 template.