Stability: are tests with implicit waits more or less stable than tests with only explicit waits?

gistfile1.txt

**QUESTION**

Some folks assert that a broad class of bugs is solved by favoring
explicit over implicit waits. We had previously been using implicit
waits by default in our code.

**ANSWER**

No substantial stability differences were observed.
Failures across chrome, ff, ie9, ie8:
  * explicit waits only: 1, 0, 2, 10
  * implicit waits only: 1, 1, 3, 10

**DATA**

We conducted an experiment:
  * run all tests 10x, across 5 platforms (vista/ff, vista/chrome, vista/IE9, xp/IE8, mac/safari 5)
  * analyze errors on each platform
    * implicit waits enabled: https://gist.github.com/4347710
    * implicit waits disabled: https://gist.github.com/4333301

win/chrome:
  implicit waits: 9/10
    1 mysterious health-check failure
  explicit waits: 9/10 success
    1 missed click

win/ff:
  implicit waits: 9/10 success (+ 1 infrastructure fail)
    1 stale element ref error
  explicit waits: 10/10 success

win/IE9:
  implicit waits: 7/10 success (+ 1 infrastructure fail)
    2 missed click
    2 mysterious health-check failure
  explicit waits: 8/10 success
    2 missed clicks

win/IE8:
  implicit waits: 0/10 success
    looked basically identical to explicit waits case--didn't look past first 2 builds
  explicit waits: 0/10 success
    27 'we are very sorry there has been an error'
    17 eyedeeme hangs after creating account