It seems that currently whenever a push or pull action encounters local files having the same name as those in the cloud, a checksum (MD5) is calculated over the local copy and compared to that stored for the file on Google's server to decide if the files are actually identical. This is certainly the safest thing to do, but it's my opinion that the default mode of operation of calculating an MD5 over every single identical file every time the user calls a push or pull is unacceptable performance wise.

I'd like to suggest that drive takes a similar approach to how rsync works and turn checksum comparison off by default. Correct me if I'm wrong, but I think rsync looks at file names, file sizes and modified times (by default) to decide if two files are identical which is probably enough for 99.999% of the time and is extremely fast. Only when the user explicitly asks for a checksum comparison (by using the -c or --checksum option) does rsync compare files by calculating checksums.