It looks like S3 supports rename now -- well via no bandwith charge copy.

   http://developer.amazonwebservices.com/connect/ann.jspa?annID=360
   http://docs.amazonwebservices.com/AmazonS3/2006-03-01/index.html?UsingCopyingObjects.html

I want s3sync to keep old copies of files -- if my local copy gets corrupted I don't want to replace my backup with the corrupted file.

So I'm thinking maybe two options might be cool:

--keep-old-revisions 3

Maintain old revisions, but keep at most 3 copies.  On upload, copy the existing S3 one to file.revision1, 2 or 3, then upload the new file into existing name.

--keep-original

Never overwrite the original that was first put into S3.  On upload, check if file.original exists, if not, copy existing S3 to file.original.  Then upload new one to existing name.

Try trickle on linux (unix??) http://www.linux.com/archive/articles/61293?tid=100&tid=47&tid=13

It lets you throttle a TCP application, kind of like nice.  I haven't tried it yet, the docs say the limits are in KB not Kb.  So to keep it at 200kbps, I'm gonna try 50KBps to send 15g in 2 days or so.

trickle -u 50 -d 50 s3sync.rb ....

[After some minimal testing, the numbers coming out of ntop do show it works but you'll have to tweak the trickle inputs to get the desired output.   On my system -d 20 -u 20 is reported as roughly 200Kbps in ntop.  50/50 was reported as 500Kbps]