difference bwtween redis force failover and automatic failover caused by a master down

Question

I want to use go-redis NewFailoverClinet as a client to set data into my database(redis-3.2.6) in a loop.

Result shows using command sentinel failover <MasterName> to trigger failover could cause data lost, but log indicates all data is saved with OK response.

Using kill -9 <masterPID> to cause a automatic failover returns a correct log which matches perfectly with data saved in database

I want to know the mechanism of failover both triggered by command and master down