https://cs.chromium.org/chromium/src/third_party/WebKit/Tools/Scripts/webkitpy/tool/commands/rebaseline.py?l=356

This step is unbearably slow when the number of tests is big (say 1000). There are two similar 3-layer for loops, but I don't think the total number of combinations is anywhere large. Likely some function calls are expensive. If the cost is unavoidable (which I think is unlikely), we should run this step in parallel.