On 20 Oct 2026, on the Havířov/Karviná/Orlová sheet, the owner asked: "Is the lines the best schedule that fits my intent (not necessarily the scripts)? Or can you do a version that fits better". A hand review found artefacts of the shared fit:

• merged variant minutes made the exactly hourly 417 approximate;
• the sparse-band fallback gave every 120 to the hourly 409 and 411;
• a minimax slot rule set 504 at :01 instead of its real :06;
• daily timetables were exact for sporadic lines (418, 420, 513).

Why: the owner cares about the schedule that reproduces the real operation, not about what the algorithm outputs.

How to apply:

• Before publishing, dump each line's real departures at its peak per direction, plus how its buses chain their trips, next to the suggested rows, and judge every approximate row.
• Hand-set (line_overrides) wherever the main pair alone fits exactly, or where one bus runs a fixed sequence. Use a daily circuit only if each leg sends the same number of buses.
• Hand-set legs follow the regular run, and odd trips are listed as off the pattern on the To check list with any slower option.

Reference: pipeline/havirov-karvina-orlova/README.md, section "Hand review". Related: every-stop-covered-rule, schedule-both-ends-same-rate, model-line-at-its-peak.