There is a race condition in InnoDB startup. A number of fil_crypt_thread are created by fil_crypt_threads_init(). These threads may call btr_scrub_complete_space() before btr_scrub_init() was called. Those too early calls would be accessing an uninitialized scrub_stat_mutex.
The bug was introduced in MariaDB 10.1.10 when the btr_scrub_init() call was moved after the fil_crypt_threads_init() call. That change was made unnecessary in a MDEV-12052 related clean-up in MariaDB 10.1.24.
The bug was caught in 10.2 thanks to the assertion that was added in MDEV-13485.