I can argue that disaster started mid-4.6, when they started juggling with rate limits while hitting uptime problems. Great we have some healthy competition and waiting for the next move from Deepmind.