These are the current coefficient ratios:

=== Cost Coefficient Ratios ===
  cpuFactor    = 1.0 (0.000219ms)
  seqIOFactor  = 762.2  (0.1666ms / 0.000219ms)
  randIOFactor = 4121.2  (0.9007ms / 0.000219ms)
DONE! 

These ratios seem off when comparing to different systems such as Postgres, where the ratios are 1 – 100 – 400 respectively instead. We think this is partly because we're using a simple collscan as the calibration query for cpuFactor instead of an in memory join.