Yes, the writes to flash memory is my concern; wearing out the flash memory. But I’m thinking that “doing a lot of development naively” isn’t as much a problem as when you’re running Linux. Unix is a disk based operating system and it historically has written to the disk willy nilly. Syslog is writing regularly, inodes are being updated when files are accessed, etc. My idea is exploring the BBB as a low power unix server.
Sounds like fun. The idea of being able to run a swap partition in this instance sure does seems nice. In fact the drive I used to test the material for the article I wrote: I forgot to add a swap partition. So I created a swap file with dd . . .it seems to work just fine.
For what it is worth. For my own ideal development environment I prefer to netboot. Simply because I have the added horsepower of my “desktop” system, and the ease of moving things around without having to interact with the BBB directly. Generally this means that single non system files I can copy over on the fly and test, while system directories or files can be done on the fly, but will require a reboot. Plus the added benefit of using “non destructive” media.
Have fun with your exploration 
I started by booting the BBB from an SD Card with one of Robert Nelson’s
images, then moved to booting from a USB stick (by copying the file system
image across and updating uEnv.txt). The stick went read-only after a few
days, and I looked at a portable USB HD - I can’t remember if I got it
booting or not. In the end it seemed easier to fetch the kernel via TFTP
and then boot via NFS - again, it was just a question of updating uEnv.txt
An NFS-loaded kernel makes cross-compling and general messing around a lot
easier, it’s reasonably quick, and I wouldn’t develop stuff any other way.
However, while setting up TFTP and NFS you must monitor the BBB console
(FTDI) during startup - there are too many details in the set-up to get
everything right the first time.
I access the BBB via ssh across the network; mostly it boots correctly just
by powering it up, but one time in ten or twenty it hangs. This problem
prevents unattended operation, and I haven’t solved it.
Will