Open the .it file in the attached .zip ("IT panning test.compat.it"):

1. There are three channels, with the following initial pan values: 64, 192, 64
2. There are two instruments, the second of which uses "Set Pan"=32 and also a pan envelope. The first instrument uses neither of these pan settings.
3. There are two patterns, the first (0) uses only instrument 1 on all three channels, and the second (1) uses only instrument 2 on all three channels.
4. The patterns play in order 0 - 1 - 0.

Impulse Tracker plays this file as follows (audio file is included in the attached .zip):

• Pattern 0 (first time): the instrument 1 (piano) notes are panned to the left and right according to the channels' initial pan values.
• Pattern 1: the instrument 2 (strings) notes are panned around based on the pan settings specified by the instrument.
• Pattern 0 (second time): the instrument 1 (piano) notes are panned to the left and right according to the channels' initial pan values.

When played in OpenMPT, however, the second instance of pattern 0 plays with centered pan on all channels. So, something about using an instrument with its own pan settings prevents later instruments from playing at the channel's initial pan value.

This test was scaled down from "Over the Clouds" by Deansdale. I have observed similar panning errors in other IT files as well, though I haven't tested yet to see if its the same cause. A couple that I'm aware of off the top of my head include "Nearness" by Bibby and "'You Did', said I" by (M)Rated. I can investigate these further when I get a free moment.