Whoa. There are various implementations of RAID-1. It is quite normal for them to read independently from both drives: when reading, one drive seeks to one place and the other somewhere else. It would be a waste to read the same thing from two drives just to verify: the checksum is much better mechanism to detect bit errors in terms of economy and performance.
But with writes, it is not the case. The writes are done in parallel in most implementations. The server had a power outage and is now starting up - it needs to quickly come up with a coherent data (data from one point in time, and it would be great if it was the most recent point in time possible). How can the server decide it quickly, if there were recent writes scattered everywhere on both drives?
And if a drive fails? The server would loose some chunks of data. We have a data loss, but worse we have incoherent data, some chunks are recent, some are old. How can the server now "roll back" the chunks from the working drive to go back to a coherent point in time? Where can it find the old version of these chunks of data?
So it's much simpler to write in parallel.
PS: There are mirrors that don't write in parallel, in the most typical usage they are used as (non-hardware) subcomponents of things like: asynchronous replication, distributed write cache, distributed filesystem, etc. Typically they are costly and used in enterprise environments.
1
Our sister site SF also has a nice explanation: http://serverfault.com/questions/339128/what-are-the-different-widely-used-raid-levels-and-when-should-i-consider-them
– Hennes – 2012-11-20T15:10:30.277