Solving, stop-and-keep, and repair
Solving is an asynchronous job: the solver searches for progressively better schedules and persists each improvement - you can watch it work, keep the best so far, or cancel outright.
While a job runs
The banner shows the live improvement counter and the current best schedule on the grid. The generation actions carry a Speed selector - Quick answers in about a minute, Standard uses the configured limits, Thorough searches up to twice as long (capped by the server's configured maxima). The choice applies to match placement, not the staffing pass.
Stop and keep ends the search and adopts the best found; Cancel discards everything new. A finished-early solve may report it hit the time limit before proving optimality - the schedule is still usable.
When a solve fails
Failures are translated into plain sentences before they reach you: which matches could not be placed, which rule blocks the schedule, and what to relax. Fix the named cause - usually capacity, a hard rule, or missing painted availability - and rerun.
Repair
Repair re-solves around the parts of the schedule you want kept: pinned or fixed placements and locked staff stay put while the solver re-places the rest. It is the right tool after late changes - a team drops out, a field becomes unavailable - because it disturbs as little of the existing plan as possible.
Versions keep you safe
Every solve writes a new version; earlier ones remain on the Checkpoints tab for comparison or rollback, so experimenting with solve settings never costs you a working schedule.