As per
https://github.com/c0pperdragon/Amig...ideo/issues/40
I used 74HVC series logic chips, which are half the speed of the 74LVC logic chips. This seems to have resolved the sparkles artefact.
Even on a relocator adapter, it has nearly entirely eliminated it - adding a 47pF cap between PiCLK and GND seems to have reduced it even further.
However, I'm still interested in finding out why X-Copy Setp 92 disk images seems to cause more sparkles compared to anything else. I also noticed that DiagROM v1.2.1 was showing a lot of sparkles too.