GPX exports from GIS information are difficult, because a normal dataset - made of a number of separate line features - does not specify the order in which lines should be interpreted.
A good GPX track has an order to it. It's designed to represent an actual journey (direction, order, etc are inherently part of the format). It may have only one 'track' or separate track 'segments'. If it does have segments the expectation would (often) be that one segment follows from another in some logical spatial sense - perhaps the end of one segment is followed by the beginning of the next nearby segment implying a break in a journey, but not a teleportation to another location.
Exporting from GIS into a GPX track, at least when using QGIS, typically introduces unwanted track segments - and may not present these segments in the GPX file in a logical order. Devices interpreting a GPX file may be unforgiving, and avoiding such issues is important.
I often hit issues with GPX export. For example, to avoid creating separate track segments it's important to snap lines, then combine them - but issues like duplicate nodes cause problems. Line direction also matters. As does any crossing of lines (entirely acceptable in GPX - less so in GIS).
Is there a reliable workflow (perhaps involving other open source software too) which will reliably produce a tidy GPX track without illogical track segments? This should be able to cope with a starting point with standard GIS line data - multiple features- perhaps not correctly snapped at line ends.