Dusk tightens right after the update, when the chain is already living in the new rule set and your flow is still proving the old one.
Nothing dramatic happens on the surface. DuskDS keeps producing finality. Committee attestations still land. The network isn’t “recovering.” It’s operating. That’s what makes people argue about what they’re even seeing.
The first failed call looks fine until it doesn’t.
Same method, same inputs, same account posture. It reaches for a condition that used to pass, and the condition is gone. Not “deprecated later.” Gone now. The attestation certificate you’re holding is real, signed by a real committee, scoped to that committee window, but it’s bound to the execution context that just got replaced. People keep saying “but it finalized,” like finality is a universal adapter. It isn’t, not when the rule hash moved underneath the path you’re calling.
So you get the annoying symptom: one transition won’t ratify while everything around it does, and nothing looks sick enough to blame.
Ops runs the usual loop. Validator uptime. RPC. Did the update propagate. Did the committee window rotate. All green. Still stuck. Someone posts the tx hash again. Someone else posts it again, but with a different explorer link, like that changes what the chain is evaluating.
I’ve been in the middle of one where the disagreement was literally minutes. “It hasn’t switched yet.” “It switches at the next slot.” Nobody wanted to be the person who says “we missed it.” Then someone pulled the on-chain parameter change and the timestamp was already behind us by one block. That was enough. Our proof wasn’t invalid. It was just for a context that no longer applies.
No grace block. No soft landing. If the execution context is stricter now, your call doesn’t get a courtesy pass because it was queued earlier. The committee can attest to what happened under the old rules. It won’t attest forward into rules that aren’t there anymore. And you can feel the room reach for a workaround that doesn’t exist.
You don’t resubmit “as is.” You rebuild. New proof material, bound to the current rule set. Same intent, different evidence. Until that happens, the chain keeps closing everything else and leaving this one path sitting there, quietly wrong at the only moment that counts.
People still ask for “just a few blocks” of overlap. They keep asking after you answer. Meanwhile the stuck call is still back there, holding yesterday’s context like it can be negotiated into today. #Dusk
