> I usually interpret 'selfish' to mean genes that produce defectors (ie no flip) since there is a local benefit to the organism (and thus the gene itself over short timescales) but Dawkins does use it a bit more broadly.

Dawkins seems to define "selfish gene" as "one caring about its long term survival and spread", whether individual tortoises like it or not. Because that's what ultimately counts - contemporary tortoises are shaped by genes which survived millions of years by whatever means happened to be successful, which needs not to be defection.

And, btw, you may want to actually read TSG, it's pretty cool :)