R survey package: svyby + svymean: one vs many variables

Question

Let's assume the a data set mydata with the variables foo1..foo20 which are factors with the labels "Easy" and "Difficult". Now let's consider this code:

library(survey)
svd <- svydesign(ids = ~ 1, weights = ~ weight, data = mydata)
svyby(~ foo1, by = ~ group, svd, svymean)$foo1Difficult
svyby(~ foo1 + foo2 + foo3 + ... + foo20, by = ~ group, svd, svymean)$foo1Difficult

Are the results supposed to be identical? Is there a reason why the results could differ? Why does it make a difference whether I iterate over each variable or use all variables at once?

Answer 1

As @AnthonyDamico pointed out, the difference was caused by NAs.