We've run into an interesting technical issue on one of my clients' projects. I am hoping someone might help me shed some light on what's happening.

We have an n-tier app with various VFP and VB-based in-process DLLs. We typically run under MTS, but we see the same behavior even when we operate outside of MTS (and, regardless of how distributed transactions etc are set up).

The issue is one of intermittent freezes -- specifically, an automation error creating the object: good ol' error number -2147221231 ("ClassFactory cannot supply requested class"). The error takes a few hours each day to duplicate, but it occurs frequently.

Here's the weird part -- at the point where the system freezes, we can duplicate the error (until we reboot the server) but only in cases where a VB component calls a VFP component. In other words...
1) VFP component returns "hello world" --> no problem
2) VB component returns "hello world" --> no problem
3) VFP component calls VB component, and then returns "hello world" --> no problem
4) VB component calls VFP component, and then returns "hello world" --> Error!!!
... it gets weirder...
5) VFP component calls VB component, which then calls VFP component, and then returns "hello world" --> no problem.

Any insights? Re-writing the app in just one language is not a real option at this point -- there are hundreds of components in each of the two languages. We are using VB 6.0 SP5 and VFP 7.0 SP1. The components in the scenarios above are simple one-to-five line processes with CreateObjects() and Returns. Nuthin' fancy whatsoever.

We're considering writing a VFP wrapper around the cases that fit scenario 4 above, which isn't terrible except for the fact that it feels like a terrible kludge and I worry about what other issues we might be masking by doing so. Then again, the evidence seems to indicate that if we start the thread with a VFP component, the issue disappears.

Thanks in advance for any help.