Review use of burstable cloud machine types for Redis Sentinel
We currently recommend the use of small burstable instances for Redis Sentinel nodes in the Reference Architectures (10k and up).
The reasoning behind this is Sentinel requires very little resources and the smallest non burstable machines on the cloud providers are notably bigger and overspecced for what they need.
However generally we always want to avoid burstable instances as they can give inconsistent CPU behavior. Task is to evaluate if this is viable as an exception for Redis Sentinel or to just recommend larger but non-burstable instances.