Fable 5 with Opus 4.8 subagents vs. Opus 5 subagents

Hey, I know I'm preaching to the choir, and this has been talked about before, but want to share my experience in an attempt to be another voice crying out to Anthropic to fix this. Like others, after initial success Opus 5, I started experiencing issues. Using it alone, I felt like I was managing an incompetent developer who constantly missed details and fail to follow instructions. So, I set Fable 5 to use Opus 5 subagents and noticed an interesting trend. Rather than me managing the incompetence, Fable 5 was doing it. Constant loops and redos burned through tokens, and I was wondering if it wouldn't have been better to just use Fable 5 alone. After a couple of days of this, I decided to set Fable to use Opus 4.8 subagents. Token usage has been cut in half, it's much faster, and output is considerably better. So, to me this is just an extra verification that there's something seriously wrong with Opus 5. And, for full context, I've done everything possible to follow Anthropic's recommendations about working with Opus 5. I won't be using it again until anthropic addresses these issues.