This happens sometimes when using Gretl.

I use the same data every time in my class and sometimes when I go here after I have restricted the sample to 1980:4 and try to forecast the next 4 years, it only lets me forecast 1 period ahead. I cannot set the End to 1984:4

Well, you didn't mention whether you also use the same model.

Obviously, if gretl's behavior is non-deterministic for exactly the same circumstances, it would be a bug. Given how computers work, it's also relatively improbable.

So my wild guess is that sometimes you have an exogenous variable which makes a multistep dynamic forecast impossible.

cheers

sven