Trigger frequency limitation
This issue addresses the problem of data arriving late and therefore filling the wrong scan iteration. When running on strip staves, we noticed that this problem gets worse as the number of hybrids (i.e. Front Ends) increases. Decreasing the trigger frequency mitigates the issue.
The first plot below shows the maximum trigger frequency that can be run without significantly affecting the s-curves as a function of the number of FEs. The second plot shows the maximum frequency as a function of 1/(number of FEs)^2, fit to a linear, which demonstrates that the relationship is approximately 1/N^2.
I know there are efforts ongoing to address this problem (in particular by Alex and @ztao), but we discussed it would be a good idea to open this issue in order to have a benchmark.