Found a way to avoid/solve the problem
In our case the GS problem always occurred just after the AP picks up the GS nodding down, the kite has the intention to follow the GS, and stops doing that, adding thrust to stay on the altitude selected in the MCP.
We talked with a KLM Captain/instructor about this. The pro's hit the AP button as soon as the Localizer comes alive, even if the GS isn't alive yet. So we now do the same. When the GS is alive and approaching the middle, we lower the IAS on the MCP with 5 or 10 kts. The plane is lowering it's speed while the GS bug passes the middle and without one exception (so far) it will follow the GS perfectly.
We'll keep on testing....
On the other hand... someone suggested to set the milliseconds in the mcp.ini file at 80 instead of within the range 50-70. We did, but then the anunciators of the gear start to behave in a very strange way. F.i. when we are at the ramp, before engine startup the EGT's start to act like a crazy windmill. So we started to play with the millisecond figure simultaniously with testing the GS problem and are still searching for the right figure in OUR system. Numbers within the range 50-70 give us a better performance.