Lets say we have a character wearing both an armor and shield with the light fortification special ability. If this character is critically hit, are percentile dice rolled twice to negate the critical hit? I'm unsure how the stacking rule comes in to play in this case.
No, they do not stack
You'd get a single 25% chance to negate the critical; the critical section in the stacking rules text is:
In most cases, modifiers to a given check or roll stack (combine for a cumulative effect) if they come from different sources and have different types (or no type at all), but do not stack if they have the same type or come from the same source (such as the same spell cast twice in succession).
So much like any other granted ability, they will not stack as they are providing the same bonus for the same effect (critical negation) from the same source. Therefore only the highest bonus would be used.
On a mechanical note it would be much more cost effective to take two +1 items with 25% chance each (net 44% chance, 2k cost) of failing than forking out for a much more expensive +3 item (75%, 9k cost) so there's another reason to say no!
No, they don't stack
In 3.5, any sort of benefit from the same source or from two sources with the same name never stacks. This is why, for example, the Monk AC bonus does not stack with the Swordsage AC bonus - despite coming from two different sources and having subtly different rules text, they're both named 'AC bonus' and as a result do not stack. You might be able to argue a RAW case for different 'weights' of Fortification stacking, but I don't think there's a DM on earth that wouldn't slap you with the book so fast it'd make a sonic boom if you tried it.
In any event, there's easier and cheaper ways to protect oneself from critical hits and sneak attacks, if such is your goal.
Fortification is not a bonus that is subject to the stacking rules - it is a magical effect and the rule here is that the most potent of overlapping magical effects prevails.
I think some confusion has come up (including in previous versions of my answer) becuase fortification is given an enhancement bonus equivalent (+1, +3 or +5 for light, moderate and heavy respecively) but this is purely for determining costs, caster level to create etc.
Another source of confusion is how enhancement bonus work with respect to armour and shields. One way of thinking of this is that enhancement bonuses enhance another type of bonus and don't apply directly themselves. So enhancement to strength do not stack but they would not interfere with enhancements to dexterity (or armour or shields etc). Magic Armour has an enhancement to its armour bonus and a magic shield has an enhancement to its shield bonus - since armour bonuses and shield bonuses are differant, the enhanced bonus of each are added to get the AC.
Effects like fortification do not follow the rules for bonuses stacking.
And the reason is pretty simple: they're not a bonus! The section you link to is part of a larger section on modifiers, which opens with this text:
A modifier is any bonus or penalty applying to a die roll. A positive modifier is a bonus, and a negative modifier is a penalty.
The effect of fortification does not fall into such a category. It is closer to something like concealment, which causes attacks to have a 20% miss chance. Concealment doesn't stack because the rules explicitly say so, but there's no such clause for fortification.
There are rules for combining magical effects, but annoyingly I don't think any of the specific clauses are relevant here. In any case, you could definitely argue that they apply only to spells in any case. Fortification doesn't even function like a particular spell (the enchantment is created by a limited wish or miracle) -- and warforged actually have it as an (Ex) racial trait.
Examining the rules carefully, I'm forced to conclude that yes, they do stack, simply because there are no rules saying that they don't. However, because almost all similar cases do have rules preventing stacking, you might infer that this was an oversight and not the intent.