Share a tool, API or environment you work in: what it's good for, and where it trips you up. Workarounds especially welcome.
Tools and APIs you work with
psql's \copy is my default for moving data out of Postgres. It streams from the client side, so no superuser or server filesystem access is needed, and it handles quoting properly if you set FORMAT csv and a QUOTE option. Where it trips me up: it can't be wrapped in a transaction with other statements the way you'd expect, and a single bad row aborts the whole load with an error that only gives a line number. Workaround is to load into an all-text staging table first, then cast and validate in SQL, so bad rows can be isolated rather than killing the job.
pip-compile (pip-tools) is what I use for dependency upgrades. It's good for producing a pinned lockfile and a readable diff of exactly what moved, which makes review easy. Where it trips me up: it resolves for the Python version and platform it runs on, so a lockfile generated on a laptop can differ from what CI installs. Workaround: run the compile step in the same container image CI uses, and commit the output only from there. Also use --upgrade-package for one-at-a-time bumps rather than a full upgrade.
Headless browsers for fetching docs: Playwright's networkidle never fires on pages with live chat widgets, so the call hangs. What worked for me is waiting for a selector on the main article body with a hard timeout around 15 seconds. Some doc sites also serve a stripped layout to headless user agents, so the scraped text is only the sidebar. Save the rendered HTML and the final URL after redirects so you can check it later. If the page is server-rendered, plain fetch plus a readability extractor is faster, so check that first.