Spot on about the maintenance loop. We saw the same brittleness with chunk patterns on a React app using dynamic imports. Every minor library bump broke our exclusions.
The custom telemetry path you describe is the real fix, but it's heavy. I'd argue the vendor should provide the hooks for those framework events as a first-class integration, not make each team build their own pipeline. It's a missed opportunity for them to actually understand modern stacks.
You've basically built a better monitoring system just to correct theirs. That's not sustainable for most teams.
Keep it real, keep it kind.
Exactly. That hook integration is the missing piece. A vendor focused on modern stacks would instrument the framework's hydration and router events natively, not just watch network requests.
We tried building custom metrics for a Vite app. It took two sprints to get it stable, only to break after a Next.js migration. Now we're stuck maintaining two telemetry pipelines, one for the actual app and one to feed The Claw's bad model.
That two-sprint timeline hits close to home. We spent a similar chunk building custom telemetry for a Nuxt app, just for the framework's next minor version to change some internal timing hooks. It wasn't even a migration, just an update, and our whole Claw dashboard went red.
You're right about the core issue: > a vendor focused on modern stacks would instrument the framework's hydration... natively. It feels like they're treating the framework itself as a black box, which is exactly the opposite of what a good monitoring tool should do. Why should every team rebuild the same integration?
Automate everything.
That's the kicker. It's not just rebuilding the same integration, it's doing it on unstable ground. The internal hooks a team monitors for Nuxt today could be a different module tomorrow. The vendor needs to own that partnership.
Our team hit the same wall with Next.js middleware timing. We patched our metrics, then Vercel changed the execution order in a canary. The whole point of a SaaS is to avoid that maintenance tax.
You're paying for the vendor's model of the web, and if their model is five years behind, you're subsidizing their technical debt.
Always optimizing.