How to avoid reshape/reshape2 melt.data.frame conflicts in R package

Question

I use melt.data.frame from the reshape2 in a package. For reasons I cannot control, I've had users load my package (which imports reshape2)...then load the reshape package, causing future errors from my package (an official Warning...and incorrect output for my code).

I've seen this discussion: Reshape package masking preventing melt from naming columns

So I know I can use reshape2:::melt.data.frame to prevent the conflict, but that generates a Note on check that I'd rather avoid. The gather function in tidyr doesn't solve it either. Here's a reproducible example:

library(reshape2)
df <- data.frame(x=rep("a",4),
                 y=c(1:3,NA),
                 key=c("g","g","c","d"),
                 stringsAsFactors = FALSE)

melt(df,id.vars = "key")

library(reshape)
melt(df,id.vars = "key")
reshape2::melt(df,id.vars = "key")
reshape2:::melt.data.frame(df,id.vars = "key") #No problem
reshape2::melt(df, measure.vars =  c("x","y"), 
               variable.name = "variable", 
               value.name = "value", na.rm = FALSE)

library(tidyr)
gather(df, variable, value, -key)
gather_(df, "variable", "value", c("x","y"))