Looking at our file-open performance data: https://docs.google.com/spreadsheet/ccc?key=0Aras0diokeHxdEc5RGtOeVI0V0xGU3FPUXBuX3ZYTlE#gid=5 There was a dramatic drop in the recorded time to open large files when we cut over to brackets-shell. There's now almost no difference between different file sizes. That seems a little suspicious given that most disk operations have seemed slightly _slower_ in -shell. I'm wondering if the greater degree of asynchronicity could be exposing a bug in where we've placed the instrumentation. Seems worth investigating.