OVH, specifically. It looks like this was it - server went down ten minutes after their maintenance window started. Of course they didn't email us about this - fortunately I have monitoring set up - and it certainly didn't just affect the API. Quebec is our biggest server in North America, so so impacted full-size images for much of the Americas as well as its own locality until I tweaked the configs.
OVH, specifically. It looks like this was it ( http://status.ovh.com/?do=details&id=13955 ) - serve