How are writes managed in Spark with speculation enabled?

Question

Let\'s say I have a Spark 2.x application, which has speculation enabled (spark.speculation=true), which writes data to a specific location on HDFS.

Now if the task (