When calling rbind on 2 data sets with common and new variables, having "labelled" variables in one of the data sets generates an error in rbind. Interestingly, the order of data sets does matter. The functions spss.get in Hmisc generates such variables.
require(data.table);require(Hmisc)
## Generate a labeled numeric
x <- c(1:2)
label(x) <- c("a")
## Generate two data.tables with 1 overlapping and 1 new column x
d1 <- data.table(v1=1:2,x=x)
d2 <- data.table(v1=3:4)
#does not work!?
rbind(d2,d1,fill=TRUE)
#this order works
rbind(d1,d2,fill=TRUE)