...alright...going to try a new SIIG FireWire 800 PCIe card (w/ TI chipset) and a new FW cable.
 
I doubt it'll do anything to fix it - but figured what the heck...and might as well get the best FW 800 card available anyway...

Choinga, I hope the new card fixes it for you.
 
I did try Firewire with my UCX at one point just to test, but I got slightly higher latency with Firewire so I didn't stick with it.
 
I haven't read all the way back in the thread, so excuse me if you've already gone over this...
on 2/27/2014, a new RME driver set was released...
http://www.rme-audio.de/en_downloads.php?page=content/downloads/en_downloads_driver&subpage=content/downloads/en_downloads_driver_fireface
 
There is also a new version of DigiCheck.
http://www.rme-audio.de/en_downloads.php?page=content/downloads/en_downloads_driver&subpage=content/downloads/en_downloads_digicheck
 
I did get a TI Card (I spent $30 USD) for the firewire test.
 
This is a reach, but can you directly power your device? Are you powering from Firewire? Is it possible that load on your power supply is impacting the device? If you can plug the thing into the wall, I'd recommend that as a trouble shooting step. I know my UCX runs without wall power using FireWire, but I do choose to power from the wall. It will not run via USB power, but again, I power from the wall.

thanks gswitz...yea, I have the latest drivers.  I don't even know what that DigiCheck is - never used it.  Not even sure it works with the FF800.
 
Well, I started the loop at 5:20PM CST and it's now 8:30PM so it's the longest it's ever gone without dying.  I'm sure it will at some point but if I can keep it up for 3 hour stretches that shouldn't be a major problem for me.  I still need to test it in a 'real world' situation...stopping/starting, recording, etc...

FWIW I tried a test looping a fairly heavy project using a UFX via USB. It looped for over an hour with no disconnect issues and no issues. I then cranked the latency down to 96 samples and did the same test. I did get a dropout a few times but no disconnect messages. So its definitely not a systemic issue. I haven't tried firewire however.

...well, I'm now at over 5.5 hours playing that loop and no dropout thus far.  Definitely a record.
 
To recap, I swapped out a Syda PCIe 1394 card for a SIIG PCIe 1394.  http://www.amazon.com/gp/product/B001EQCNMW/ref=oh_details_o01_s00_i00?ie=UTF8&psc=1
 
...and I swapped out my cable for this one:  http://www.amazon.com/gp/product/B0081GL734/ref=oh_details_o01_s01_i00?ie=UTF8&psc=1
 
I still haven't done any practical testing but with my prior setup I never went this long without the dropouts/error messaging popping up showing the device was disconnected.
 
All my testing was down at 64 samples, but my project is exceptionally light - one track with me talking for 15 seconds or so...no FX...and looping.  So 64 samples shouldn't be an issue at all.

...well, it played the loop all night long - like over 8 hours. :)
 
So, who knows either my Syda card or cable was jacked up or the RME drivers don't like them anymore or any other combonation of craziness.  Either way, seems the new card + cable fixed me up.
 
Very strange...

That  conclusively proves it was a hardware issue. If nothing the disconnect notification helped alert you to the real underlying problem :)

Choinga,
 
I totally recommend installing DigiCheck and exploring it.
 
There tons of great functions like an awesome Frequency Analyzer, recorder and other fun things...
http://www.rme-audio.de/en_support_techinfo.php?page=content/support/en_support_techinfo_digicheck
http://www.rme-audio.de/support_techinfo.php?page=content/support/en_support_techinfo_totalyser
 
You can record a show using the record feature. It records all tracks to a single file and it is amazingly efficient. Then, after the show, you can extract all the data to separate wav files.
 
It counts as my favourite way to capture a live show.
 
Smiles.

Noel Borthwick [Cakewalk]
That  conclusively proves it was a hardware issue. If nothing the disconnect notification helped alert you to the real underlying problem :)

looks that way.  I added 4 plugs to my original track, cloned it about 20 times and added a few heavy VST's on the master buss as well and been running it for several hours now with no dropouts or problems.  CPU is getting about 55-60% utilization and Sonar is using about 700MB of RAM...still on 64 samples, so it's getting pushed pretty good.

Glad this finally came to a resolution! Still looking forward to the sample rate change not issuing the message   :)