Task 4788 is broken

When I try downloading the task in R I get:

> tsk <- getOMLTask(task.id = 4788)
Task '4788' file 'task.xml' found in cache.
Task '4788' file 'datasplits.arff' found in cache.
Data '414' file 'description.xml' found in cache.
Data '414' file 'dataset.arff' found in cache.
Error in .jnew("weka/core/Instances", .jcast(reader, "java/io/Reader")) : 
  java.io.IOException: keyword @relation expected, read Token[=====], line 1
> traceback()
8: stop(list(message = "java.io.IOException: keyword @relation expected, read Token[=====], line 1", 
       call = .jnew("weka/core/Instances", .jcast(reader, "java/io/Reader")), 
       jobj = <S4 object of class "jobjRef">))
7: .External(RcreateObject, class, ..., silent = silent)
6: .jnew("weka/core/Instances", .jcast(reader, "java/io/Reader"))
5: RWeka::read.arff(file)
4: arff.reader(f$datasplits.arff$path)
3: withCallingHandlers(expr, warning = function(w) invokeRestart("muffleWarning"))
2: suppressWarnings(arff.reader(f$datasplits.arff$path))
1: getOMLTask(task.id = 4788)

There appear to be similar problems in python.

Start by reproducing getOMLTask(task.id = 4788) in R and inspect the cached task.xml, datasplits.arff, and dataset.arff files. Compare the similar Python failure and determine why the ARFF parser encounters '=====' instead of @relation; done means task 4788 downloads and parses successfully.