// problems solved
A few of the toughest bugs and deadlines I’ve worked through: what went wrong, what I tried first, and what finally fixed it.
The problem
1 in 12 payment confirmations went missing during a client’s flash-sale launch. Customers were charged but saw no order.
What I tried
Checked the payment gateway: its logs showed every callback delivered. Scaled the API from 4 to 12 servers: the failure rate stayed the same.
The fix
The webhook and the browser redirect raced to write the same order. Added idempotency keys and a row lock. Zero drops since.
The problem
A field-survey app had to sync 3,000 offline records over a 2G connection before the judges’ demo.
What I tried
Full JSON sync timed out after 40 seconds every time.
The fix
Switched to delta sync with compressed batches and a resumable queue. Sync finished in 6 seconds. We placed first.
The problem
A Node.js reporting service climbed to 3 GB of memory and restarted every six hours.
What I tried
Heap snapshots pointed at an in-memory cache. Capping it slowed the leak but did not stop it.
The fix
Each request added an event listener and never removed it. One-line change. Memory now sits flat at 180 MB.
// got a problem?
Send it over. I’ll tell you honestly whether I can help.