The failure your SLO allows. At 99.9% over 30 days, that is 43.2 minutes of total outage. While there is budget left you ship, and when it is gone you stop and work on reliability.
It turns an argument into arithmetic, and it applies directly to agents. If they let you ship faster than your budget can absorb, the budget says slow down, whatever the velocity chart says.
Say it to your agent
“We run at 99.9% over 30 days, so we have 43.2 minutes. Tuesday's incident was 18 minutes at 30% failing, about 5.4 budget minutes. Work out what is left from the real logs. If we have spent over half in the first week, hold the risky migration and work on the cause.”