Our settlement bot sends several transactions per minute from one address. After we parallelised it, sends failed with nonce too low and replacement transaction underpriced, and some transactions sat stuck for hours. A single nonce manager per address and correct fee bumping fixed it.
Most review comments are about naming and style, while the bugs walk past. How we review for correctness instead: follow the data, probe edge cases and error paths, look for races and trust boundaries, keep PRs small and hand style to machines.
Our order router ran fine as one process. When we scaled to two workers, a third of private API calls failed with EAPI:Invalid nonce. Kraken requires an increasing nonce per API key, and two processes generating millisecond timestamps raced each other. One key per worker fixed it.
Our webhook dispatcher threw ConcurrentModificationException a few times a day and silently skipped deliveries the rest of the time. Two bugs shared one list: removal inside a for-each loop and an admin thread editing it concurrently.
A reminder job sent 380 emails to one customer and none to the rest. The code was correct under Go 1.22 loop semantics, but the module's go.mod still said go 1.20, so every goroutine shared one loop variable. Here is the full trace and fix.