The flaws were discovered in August by researchers at Wiz, OpenAI, and Calif, according to SecurityWeek. Both come down to how PHP cleans up after itself.
When a request pushes PHP past its memory limit, PHP stops abruptly. In that rushed stop, some of its internal notes get left half-written or pointing at space that has already been handed back. A carefully built request can take advantage of that mess and get the server to run the attacker’s own instructions.
Back to the restaurant: imagine an order ticket gets torn in half during a rush. A clever customer scribbles their own instructions on the leftover half, and the next cook follows them without question.