Thanks bingo600 for the clarifications.
According to your information I assume the GPIB is an event triggered bus and the traffic there is generally very low, so it is very unlikely to hit the maximum bandwidth limit, at least with the 34401a.
Now let´s discuss what I read about GPIB.
I heard that it is possible to send over the GPIB screenshot of Spectrum Analyzer [SA] measures, in this case while the image is being transfer the GPIB is 100% busy (no other communication possible).
Now imagine I was polling cyclic every 10NPLC a Voltage measurement with my 34401a on the same GPIB Bus, at the time the SA image will be sent I think I will lose some measures.
Same story with a very slow (I imagine, still have to test) 1985 device like the Keithley 220. I trigger the 220 and I hope it will give me back what I need before I need to get the data from the 34401a.
Can you confirm me the GPIB is a "one talker only" bus? If yes, then the bus performance is affected by:
1) Number of talker devices
2) How fast they can talk
3) How much they have to talk
Again, I have just started with GPIB and I say thanks to everyone in advance who is explaining me how it works.