Amazon ElastiCache exists to keep load off your database, which means your database was never sized for the traffic the cache absorbs. That gap is the entire risk, and it is one of the most reliable ways to turn a minor incident into a major one.

A cache outage becomes a database outage in four steps:

1. ElastiCache becomes unreachable, and every request that would have been served from memory falls through.
2. Your database receives its normal traffic plus everything the cache was absorbing, arriving at once.
3. Query times climb, so requests take longer, so concurrency rises, so query times climb further.
4. The cache recovers in minutes. The database takes far longer, and the incident review blames the database.

Gremlin lets you measure that gap safely. Blackhole experiments remove the cache endpoint from your application's view while you watch what reaches the database. If the answer is uncomfortable, the mitigations are well understood: request coalescing, stale-while-revalidate, a circuit breaker in front of the database. All of them are easier to build on a Tuesday than during the outage that proves you needed them.