Cross-browser testing — how many browsers do you actually test against?

Theoretically we should test Chrome, Firefox, Safari, Edge. In practice: - Chrome is 65% of our user base - Safari is 20% (mostly iOS Safari which is different from desktop) - Firefox and Edge together are ~15% Running our full Playwright suite against all 4 browsers would 4x our CI time. Current approach: Chrome only in PR pipeline, Chrome + Firefox + Safari in nightly. Is that the right split? Any shortcuts for Safari testing specifically (since you can't run true Safari on Linux CI)? Also curious: has anyone ever caught a real production bug that only affected Firefox or Edge in 2026?