Code review comment for lp:~mbp/launchpad/less-randomness

socratic dialog:

<wallyworld_> poolie: isn't the whole point of using a rng to provide different input each time?
<poolie> wallyworld_, so, generally, my experience is that using a rng in tests is a bad idea
 if there are some random values that catch a bug, which is possible
 there are problems
 one is, the particular values are likely to come up only very rarely, so the bug may be latent for a long time
 secondly, it will be hard to reproduce it, so perhaps hard to understand why it's failing
 oh, and also, it may fail at inconvenient times
 for example on buildbot but not ec2 etc
<poolie> wallyworld_, so this particular test is testing code that itself uses the rng, so i think it has a valid reason to touch it
 but it should isoltae itself
<wallyworld_> poolie: that all makes sense. thanks for explaining