You've probably checked this, but when looping, your clips MUST be an exact length in measures.
 
A tick or 2 off in either direction will mess with the timing

Yep, looped exactly on the dot. 

Is your timing master in your audio settings the same pair you are using for your main outs?  I have had some funky issues when it's set to another output pair.

Silicon Audio
Is your timing master in your audio settings the same pair you are using for your main outs?  I have had some funky issues when it's set to another output pair.

Yep :)

James P
 
scook - thanks, I tried your suggestion of switching on the metronome and if I do that, they do seem to stay in time with each other. Strange but not really something I can (or should have to) do in a real world example.

If it solves the problem why not use it? Mute the metronome bus if you do not want to hear it.

scook
If it solves the problem why not use it? Mute the metronome bus if you do not want to hear it.

You're right of course. Don't know why, I just didn't think of muting the metronome bus. I will do that and thanks for the heads up (how did you find that out BTW???)

But this is typical of the sort of bugs I keep coming across that I have to stop and ask for help on. It's stupid that I should have to enable the metronome but then mute its bus in order to fix a timing bug. Do you know if this been fixed in X3?

The subject comes up every now and then (sometimes twice in the same day http://forum.cakewalk.com/FindPost/2921460) and the solution has been around just as long. Now that you know, it is a simple matter to never turn the metronome off, just mute it. 
 
I have not installed X3 so have no first hand experience with it but the link above reported a similar issue in X3b. Not sure if I would have tested looping with the metronome off since muting the metronome is part of my workflow.

scook
The subject comes up every now and then (sometimes twice in the same day http://forum.cakewalk.com/FindPost/2921460) ...

That's funny - it is the exact same issue and from the look of that post, it is still happening in X3. Anyway, will make sure I make it part of my workflow too from now on. Thanks again - good catch!

Incidentally if anyone's interested, I've just noticed that this also helps with a related problem I was having where some 32 bit VSTi's cause a click when playing back a looped section. You have to install jbridge (as opposed to bitbridge) and then do the enable & mute metronome and the clicking in the looped section disappears. So obvious really ;) 

Yes it's been reported against X3 by a few people already, but it makes me crazy seeing all these complaints about it because I have never been able to reproduce it in any meaningful way. The "best" I could do was a sync error of 40 milliseconds after 180 iterations (!). And I could not get it to happen with loop boundaries on beats. Also it went away when I switched to 44.1kHz from my usual  48kHz.
 
Some others have reported it went away when they switched from WDM to ASIO driver mode, so that's worth trying. I checked out WDM driver mode with my first couple interfaces that also supported ASIO, and WDM exhibited sync issues when restarting playback in the middle of a project, so I went with ASIO and never looked back.
 
Bitbridge's adding an extra empty buffer of audio is another issue, but it shouldn't be an issue in this case since Stylus is available as 64-bit native plugin.

brundlefly
Also it went away when I switched to 44.1kHz from my usual  48kHz.
 
Some others have reported it went away when they switched from WDM to ASIO driver mode, so that's worth trying. 
 
Bitbridge's adding an extra empty buffer of audio is another issue, but it shouldn't be an issue in this case since Stylus is available as 64-bit native plugin.

Before trying scook's solution I tried just about everything and it made no difference to me - 44.1/48khz as well as WDM/ASIO. The metronome solution definitely fixes the drifting for me - hope the bakers make a note of that as it does seem to affect others as well.

The jbridge/bitbridge thing wasn't actually Stylus-related (which as you say is 64bit) but a different 32bit VSTi - I only mentioned it because I noticed it was also fixed by the metronome (in conjunction with using jbridge).