Consistency is a good thing, but it can be outweighed by other considerations. If my choices are between consistently giving the answer '2 + 2 = 5' on a test or sometimes giving '2 + 2 = 5' and other times ' 2 + 2 = 4', the latter is probably preferable. Kaj's argument is that if you core goal is EA, then spending hundreds of thousands of dollars on cryonics or heart surgery is the normatively wrong answer. Getting the wrong answer more often is worse than getting it less often, even when the price is a bit of inconsistency or doing-the-right-thing-for-the-wrong-reasons. When large numbers of lives are at stake, feeling satisfied with how cohesive your personal narrative or code of conduct is is mostly only important to the extent it serves the EA goal.

If you think saving non-human animals is the most important thing you could be doing, then it may be that you should become a vegan. But it's certainly not the case that if you find it too difficult to become a vegan, you should therefore stop trying to promote animal rights. Your original goal should still matter (if it ever mattered in the first place) regardless of how awkward it is for you to explain and justify your behavioral inconsistency to your peers.