You have semi-synchronous replication configured and working with one slave.

No committed transactions are lost.

The slave automatically identifies that the master is unreachable and performs any required actions so that applications can start using the slave as the new master.

As soon as the incident happens, application can read data from the slave and rely on it to return a full and current set of data.

Reads from the slave can return outdated data until the value of the rpi_semi_sync_master_timeout variable is reached.

Reads from the slave can return outdated data for some time, until it applies all transactions from its relay log.

A small amount of committed transactions may be lost in case they were committed just before the disk failure.