Idk if calling it "horrible" is very fair. Forecasting details beyond 3 or so weeks is beyond the ability of any model. In fact, one of the big realizations of chaos theory was the fact that it's impossible to produce an accurate deterministic forecast beyond about a month, even
if our measurements of the atmosphere were very accurate. Like
@SD says, after a certain point the ensemble members are so noisy that they stop reflecting the details of the model's initial conditions and start reflecting the steadier, larger-scale forcings/patterns (things like ENSO).
To that end, the weeklies do a pretty decent job. If anything, this just reflects how lucky we are to be eyeing such a favorable pattern during a La Nina.