Reports of a mythical Facebook Phone have surfaced over the past two years without a real device ever in sight.

Now AllThingsD is the latest to rekindle rumors of such a device. Liz Gannes and Ina Fried reported yesterday that Facebook has picked HTC to build an Android-based phone emphasizing HTML5 for shipment in 12 to 18 months. The company’s chief technology officer Bret Taylor is overseeing the project, they report.

But the project itself has had a long and troubled history internally at Facebook, say sources with knowledge of it. Originally a skunkworks project that was kept secret from employees for months, Buffy is indeed the codename for the Facebook phone.

Sources tell us that Buffy has been a “train wreck,” and was one of the many reasons that Facebook vice president Chamath Palihapitiya, who used to oversee the project, left the company. Several of the other people Gannes said worked on the project like designer Matt Cahill and Joe Hewitt have also moved on from the company, after spending around four years or more there and vesting. Gannes is now saying that Buffy has been reborn as a more conservative project — an HTML5-centric phone based on Android instead of a wholly integrated device with proprietary hardware and software.

But one of the many problems with forking Android, which means building on a competitor’s open-source code base, is that a Facebook phone would always be at least one version behind the latest technology. That’s fine for Amazon, which is trying to master a narrow use case — primarily reading, with some video consumption and light gameplay — at a lower price point.

A Facebook phone’s social features would need to be so compelling that the device would stand out against the very latest Android phones, on which people can already download the standalone Facebook app. On top of that, there are always risks that Google could change its APIs, codebase or implement policies to make life difficult for Facebook, just as Facebook has always done for certain developers on its platform.

Facebook also risks alienating current carrier and hardware partners, just like Google has in its deal to buy Motorola Mobility. After TechCrunch’s Michael Arrington wrote of a Facebook phone 13 months ago, the company had go to around on calls and in meetings to reassure partners with the line, “There is no Facebook phone. There are many Facebook phones.”

That’s not to mention that Fried’s suggested 12- to 18-month timetable before shipping would put a Facebook phone very much behind the curve. That would be two and a half years behind Microsoft’s launch of Windows Phone 7 and a year and half behind Amazon’s Kindle Fire.

So why is Facebook still working on it? “Internally? They don’t know what to do,” says another source with knowledge of the project.

The reality is that Facebook is in a difficult position with respect to mobile.

The company has the fortune and misfortune of being born in 2004, when it was young and nimble enough to leverage social networking into a powerful grip on the desktop web. But it was too young to influence what has now become a two-horse race between Android and iOS. Just a year after Mark Zuckerberg launched the social network out of his Harvard dorm room, the iPhone project had already begun at Apple and Google had acquired the company that would become Android.

Facebook is now operating in a world where it has the killer app with more than 350 million monthly actives on the mobile web, iOS, Android and feature phones. But it doesn’t have much wiggle room to expand that into something bigger that could contribute revenues to the company’s bottom line over the long-term.

Trying to build a platform on a rival’s platform is hard. Apple and Google are too savvy to allow a third-party like Facebook to become powerful on their own platforms. Both platforms also have rules around payments, making it difficult for Facebook to extend its 30 percent tax through Credits to smartphones.

To do it, Facebook is trying to get around Apple and Google in two ways: HTML5 and an Android-based phone with compelling social features. But both will only start to pay dividends much later. An Android-based phone is problematic for the reasons listed above.

HTML5, while promising, is too immature to give Facebook traction in the near-term. Last month, Facebook quietly launched an HTML5 platform for the mobile web plus viral channels for native iOS apps. But many top mobile developers, who are in that sweet spot of $1 to $5 million in revenue per month on iOS and Android, told us they frankly weren’t interested in diverting scarce development resources into building HTML5 versions of their games until they saw more traction. Only the most loyal of developers like Zynga, Wooga and Storm8 (which was co-founded by Facebook alums) were launch partners. That’s not to say this isn’t the right long-term choice. It probably is, but it will just take a very long time.

Facebook has also internally reshuffled mobile executives in the last few months. In addition to reportedly moving Taylor onto managing the phone project, Vaughan Smith, who handled acquisitions and corporate development, now also oversees mobile partnerships. Emily White, who oversaw the company’s Groupon competitor Deals until it was discontinued in August, also added mobile partnerships to her responsibilities. Henri Moissinac, who came to the company from eBay like Smith, is said to be no longer working on mobile business development.

Buffy “was an interesting story six months ago,” says the source. “But it isn’t anymore.”