The IO error on the devices only seems to happen on Linux not my windows box. When I moved those units to the raspberry pi it started having issues.
I can't say for sure there isn't a problem with a cable but I can hook it back up to windows 7 tomorrow maybe and try that.
I built the version from the git. I don't so far dare to stop cron from killing it every half hour. I can't watch it constantly and so far my script to start them up at boot isn't working correctly.
I do have a side issue that isn't cgminer directly. I have been plagued by crashes for a while. I haven't in the past seen anything to make me think mining specifically caused it. The most recent crash left me with the following error. I am including the last line of cgminers output and the crash report. Crash was 3.5.1 not newest from git. crashed after 57 minutes of running. 3 more and it would have restarted at the hour.
After that is some unwind kdb info I can type that up if you would like. It seems like there is a kernel error. I got lucky as for a change the monitor was still getting a signal. Usually power management manages to keep the signal off making the crashes far less useful. I don't know that this will help anything.
There's no doubt the USB in the RPi is less reliable than a real PC. Part of the problem is that the network device is actually a USB device low down meaning USB comms are compromised as a result, and it's all on the same bus.
Bus 001 Device 004: ID 03eb:204b Atmel Corp. LUFA USB to Serial Adapter Project
Bus 001 Device 003: ID 0424:ec00 Standard Microsystems Corp. SMSC9512/9514 Fast Ethernet Adapter
Bus 001 Device 002: ID 0424:9512 Standard Microsystems Corp. LAN9500 Ethernet 10/100 Adapter / SMSC9512/9514 Hub
Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
Unfortunately I can't really start digging into kernel code, but I can at least suggest trying a different distro (though I believe I've said that before). I can reliably crash my RPi by just plugging something into the USB slot that draws more power than usual (such as the BF1 stick) without mining even running. Would be interesting to see now that you've built from git whether you get complete failures with transfer errors or just IO errors as a temporary glitch now.
Due to me fat fingering an edit on cron I think if my math is correct it never restarted cgminer with the newest built from git and made it overnight.
That having been said I am not sure that */220 wouldn't ever run the restart script. I intended to put the new version from */30 to */20 for minutes making it restart 3 times an hour. Since that had stopped the hanging before.
I have moved things around and I still do get errors of the IO type. While moving things I noticed no errors with only 4 jalapeno's. As I was adding the next 4 I didn't see IO errors until the third from that set was added. After having all 9 back on I have only noticed a couple of IO errors. Maybe I should spread the proverbial love to a couple of more raspberry pi's.
Upgrading in the past hasn't helped me at all. I get more crashes and more corrupted sd cards. I tried very shortly arch but my Linux command line is lack luster at best so I went back to raspbian. I even tried MinePeon but the image I downloaded was corrupted. I really do hope the newest version won't require as frequent of a restart.
and it hasn't yelled at me to tell me I'm not at a difficulty level it likes 
and otherwise its not cell phone friendly 
.