Larry, this not really about round trip latency even though it is connected. 
This is about the way the driver looks at your input and output latency and reports this to Sonar so Sonar can apply delay compensation to the newly recorded tracks. 
 
All systems have latency in and out which added together we call Round Trip Latency
 
It sort of starts with the GUI. We are looking at a note in piano view or a waveform in track view and we expect to hear the sound instantaneously as the now time scrolls past it.
This sound has to pass through a few things before it gets to our speakers. Look at your  Audio settings in Preferences and note what Sonar  is saying your input and output latency is.
My 6i6 shows 12ms output.  So sonar actually plays the sound 12ms early so I hear it in sync with the GUI,,, but wait, my input latency is 14ms so that would put the new track down 14ms late,,, but Sonar is smart enough to use that input figure and puts my new track 14ms back so it's now in perfect sync. 
 
This is why the loop back test, it proves everything is working the way it should... if it isn't you can adjust the manual offset,,, or my self I would buy a new interface.  
 

I agree, everyone should loop test the interface when it is installed for the first time.
 
You should check alignment / offset in samples at different ASIO buffer sizes to see if it is the same (which it was in my case, i.e. this must be the interface's internal buffer size which is apparently not included in the reported ASIO RTL) and define the proper settings to compensate in Sonar, see below
 
I did it for the new MOTU AVB series a few weeks ago: http://forum.cakewalk.com/FindPost/3466624
 
Here's what I found:
 
While my Roland interfaces were always spot on (i.e. did not require a manual offset), I had to apply +82 samples offset (@96kHz) to obtain proper alignment.
 
The required offset seems to be constant across different ASIO buffer settings, so it is fine to define once and leave it (see "Sonar Preferences / Sync and Caching / Manual Offset" at bottom of tab, set to +82, keep "Use ASIO Reported Latency" checked")
 
However, there is one potential pitfall: If you change the number of channels on the AVB to be sent from/to the computer, the interface name in Sonar changes (as it contains the number of I/O in the name). Hence, you will have to remember to redefine this in case you change the AVB to use more/fewer channels.
 
 
I also recommend to loop test your MIDI devices to know what you got as MIDI latency / jitter can be bigger than audio latency if you use one of today's top audio interfaces.

A MIDI loopback test? Hmm, 
If your talking about the triggering of outboard midi gear I guess that would be easy to test, 
Just patch it back and record the audio and see if that lines up. 
 
but a midi out to midi in,  that's impossible as most people are using a USB midi input. And the controller uses it's own driver, not the audio driver. 
But I assume an interfaces midi output is controlled by the interfaces driver. 

Cactus Music
A MIDI loopback test? Hmm, 
If your talking about the triggering of outboard midi gear I guess that would be easy to test, 
Just patch it back and record the audio and see if that lines up. 
 
but a midi out to midi in,  that's impossible as most people are using a USB midi input. And the controller uses it's own driver, not the audio driver. 
But I assume an interfaces midi output is controlled by the interfaces driver. 

I was, of course, talking about physical MIDI OUT to MIDI IN connections using a cable (like in your audio loopback test) ... you can do that for any MIDI interface or USB+MIDI interface that has MIDI OUT and MIDI IN jacks (which applies to many consumer and prosumer audio+MIDI interface combos) and get an idea of your MIDI performance (and you might be surprised by how far it may be off the audio), see here: http://forum.cakewalk.com/MIDI-latency-and-jitter-for-Sonar-and-various-MIDI-devices-m3462210.aspx#3462210