Too bad major projects got stuck on 2.10 and there is no safe way to upgrade to anything newer. We were running into issues even upgrading to 2.11 with Spark.
As far as I'm concerned, they are working on a way to upgrade to Scala 2.12. But closure serialization is the main blocker IIRC, for now.
I appreciate they have legitimate reasons and am not critical of them for it. It's just that many users of Spark are doing so through vendors like Cloudera, Hortonworks, Amazon EMR etc. And so, since companies often don't have the money/time to just continuously upgrade production Hadoop clusters often users are stuck on older versions of Spark for a very long time e.g. months or years.
This translates to if you're working in Big Data everything you touch must support 2.10.