There was a post here a little while ago about a google service (I think it was 'locky') that the SREs would deliberately disable if they had over-met their SLA

There was a post here a little while ago about a google service (I think it was 'locky') that the SREs would deliberately disable if they had over-met their SLA. This forced the developers of dependent systems to handle outages correctly.