Spikes and Gaussians are related signal types, and if there's a Gaussian in the middle of spike RFI (drifting or frequency stable) it's pretty certain that the Gaussian is part of the same RFI.

We were already applying spike RFI zones (frequency stable) to other signal types as well. But we started finding examples of Gaussians in drifting spike RFI also; for example, this.

Software-wise, it was a bit tricky to do this. RFI removal is done separately for each signal type, and the areas of drifting spike RFI (triangles in time/frequency space) are computed on the fly for spikes. So what I did was:

• Do RFI removal for spikes first, and write the time-ordered list of triangles to a disk file. There are about 524M of them.
  
• During Gaussian processing, read the triangles into a time-limited "sliding window" buffer, and maintain a corresponding R-Tree of their bounding rectangles.
  
• For each Gaussian, enumerate the rectangles containing its time/frequency position, and check whether it's in any of the corresponding triangles.
  

This seems to have done the job.