How do you prevent a cache stampede when using Spring's @Cacheable annotation?

Tests whether you know @Cacheable alone doesn't coalesce concurrent misses, and what actually does (locking, soft TTL).