SLOs/SLAs in practice: how to set realistic targets

setting realistic SLOs and SLAs is proving to be a challenge on our team. everyone wants 'five nines' availability, but when you break down what 99.999% actually means (about 5 minutes of downtime per year), the engineering implications for redundancy, testing, and operational rigor are enormous. we're struggling to define what's truly acceptable downtime for different services given their business criticality. how do you go about determining the right SLO for a given service or feature? what's your process for handling SLO violations? we want to aim high, but not set ourselves up for constant firefighting and burnout because of unachievable targets.