The solver software takes a fairly conventional approach of solving a whole Nonogram puzzle by tackling just one line at a time. The intention is not necessarily to solve each line completely in one go, but to partially solve it in several turns, each time determining none, some or all of the remaining unknown cells' states.
Note that this doesn't mean that you just keep applying this operation to the same line continuously. Subsequent applications will have no further effect on the state of the line, i.e., the operation is idempotent. Instead, knowledge of the state of the line must be increased between these operations by some other operation. In practice, that means applying the operation to a perpendicular line (or occasionally guessing).
The partial solutions of one line will almost always be interleaved with the partial solutions of other lines. After a cell's state is determined while working on one line, that information is available to the perpendicular line which crosses the first line through that cell, so each partial solution fuels subsequent perpendicular attempts, until either a complete solution to the puzzle is found, or the ‘fuel’ runs out.