As some experience shows [1] [2], accident happens when people are authorized to push the code themselves (after diff review or not [1]...).
It may be times to discuss some contribution policy adaptations.

Discussion happened on irc some time ago between @zack, @nahimilega and me. Possible policies so far:

• (current) contributor can push once the diff is reviewed and accepted.
• contributor never pushes. Staff member pushes once the diff is reviewed [3]

Please feel free to amend the description ;)

[1] D1658#38146

[2] P455

[3] We'll then have to discuss the git adaptations needed (git commit message format, merge policy, etc...).