I don't think it's a given that the built-in FTL would screw up so badly as to cause a big increase in program/erase counts; this is a relatively expensive general-purpose computer, not a generic $5 USB stick. It's certainly worth investigating what the performance tradeoffs are with various hardware FTL implementations versus a sophisticated kernel-based approach. Storage performance seems likely to be a part of user experience, so one would hope smartphone designers are keeping an eye on the write amplification and latency when they make these decisions.
The idea of a "flash filesystem" that's built for low-performance applications is one I'm happy to see die. One of the big advantages of flash storage is the potential for parallel operations. I think moving to full-featured filesystems, potentially with tweaks to be friendly to solid state storage, will produce a much better overall result than you'd get by starting with a software FTL and then trying to grow a specialized filesystem on top.