It'd be nice to have a more firm benchmark, perhaps one with the "strange fruit" where the ground truth is known (which also allows for easier simulation).
And it'd be really good to compare these on EC2 with the same response rate. That could also allow comparison between NEXT and Salmon and when they start to break down.
Currently, there's a beta comparison of NEXT and Salmon on the documentation: https://docs.stsievert.com/salmon/benchmarks/adaptive.html
It'd be nice to have a more firm benchmark, perhaps one with the "strange fruit" where the ground truth is known (which also allows for easier simulation).