Further testing has shown that this is an issue for .wav files. The workaround is to convert any .wav files to .mp3 and use those in the module instead.
Note - this is only really necessary for .wav files that the module may play in quick succession, either by user input or automation in the module.
Also, if multiple mp3 play sound traits are activated simultaneously - such as by a fast-acting GKC, the mp3 sound files will be additive, increasing the sound volume in proportion to the number of pieces activated. A better solution might be for the GKC to suppress sound and play its own. This is fiddly to do in current versions, so I made a feature request for this to be part of the GKC, like suppress reporting is.