From my understanding these stats are based on the requests file extension only.
If that's correct this also includes requests that resulted in a 404 status or in a 200 status but where the actual response
probably was a styled or plain text 404 page (that did not result in a 404 status)?
Also I was wondering why .ttf and .otf are included since IE9 only has partial support for these two? I'm not sure, but wouldn't
it be safe to say that all requests that incl. wOFF or EOT can be considered as "bullet proof font-face embedding", but requests
that include only .ttf and/or .otf probably don't render anyway?
It would be great to differentiate between »real web fonts« (typefaces) and icon web fonts btw.. I'm aware of the fact that this can
only be done by filtering the most common icon fonts by name (fontawesome, glyphicons, icomoon and such). A quick test lowered
the 48% percentage in January to something like 36% for this scenario.
Thanks for clarifying: Lars