Something feels odd about Sonnet in Claude Code lately

I don't know if my take agrees with anyone but lately I've noticed Sonnet has been a bit very helpful in Claude code compared times when I'd genuinely have to switch to Opus to do some meaningful task. I know this because I'm a heavy user of Sonnet since I don't have a lot of funds like many of you have to afford $100 or $200, and Opus is also almost always out of the scope except when I see Sonnet is really failing things. And I seem to get quite the a lot of feedback submission in my Claude Code sessions and this somehow triggers me to think whether they're A/B testing sometimes or it's just normal behavior? Also for every task that doesn't seem to work out Claude seem to auto construct a feedback submission to Anthropic which I've occasionally accepted to submit. Generally, based on my experience I've always had with Sonnet, something seem to have improved in the past couple of about 2 weeks? Unless am wrong. And don't confuse my take on the fact about increased limits being reduced yesterday, it's just a general take.