A friend of mine who was once at Google mentioned that in a meeting once their SRE group was told by their senior manager roughly "you've broken stuff in $time_period far less than your outage budget, and that probably means you should've been rolling out features faster".

Risk -levels- are a choice and also a trade-off.