MultipleOutputs is very slow when trying to use an inexistent multiple output

[pereferrera]

If the user is trying to write to a multiple outputs that hasn't been configured, and the job hasn't a default multiple output defined, successive calls will become very slow. The reason is that, before checking if the multiple output name is valid, a series of expensive actions are performed (reflection, output format instantiation, ...). This can be seen in class PangoolMultipleOutputs, method getRecordWriter().