The first trials with showed that the movement of the laser head is to slow:

"Burning" a pcb layout did exactly that:

the photoresist was burnt so it stuck to the board, during the development process, instead of dissvoling as expected.