Your timeline aligns closely with what we've measured. We observed a step-change in initial response time for P2 tickets beginning roughly eight months ago, corresponding with their Q3 platform release. The median acknowledgment time shifted from 52 minutes to 285 minutes in our data.
Your examples are telling, especially the archive node communication errors. We found that even with clear log evidence appended at ticket creation, the probability of an unnecessary L1 to L2 handoff remains around 40 percent. This indicates a triage process failure, not just queue depth. The problem is their frontline can't, or isn't incentivized to, parse provided diagnostics.
A practical workaround we've adopted is to structure the ticket title as a diagnostic statement: "P2: Service X failed start on v7.5.2 - FATAL ERROR: dependency Y mismatch". This has reduced our tier-bounce rate by about 15 percent, though it shouldn't be necessary. Have you seen any correlation between specific patch versions and particularly egregious delays?
Latency is a liability
Yeah, that 4-6 hour P2 acknowledgment window is rough. I've found that routing issues get worse when support teams are siloed from engineering's deployment cycles. It reminds me of when our Argo CD sync failures would get stuck with L1 because they didn't have the rollout manifests.
The title trick someone mentioned does help. I go a step further and link directly to a specific line in our hosted log file in the initial ticket. It seems to cut down on the back-and-forth for service start failures. Still, you shouldn't have to build the runbook for them.
git push and pray
The log link trick only works if their agents are allowed to click external links. Found out the hard way that some locked-down support portals strip them out, which adds another loop of "please attach the log file directly."
Your Argo CD example is spot on, that exact scenario is why support silos kill response times. The L1 team's runbook is probably two years out of date and doesn't include anything from the last three major platform updates. So they default to bouncing it.
Have you seen any improvement since they started that new support portal, or is it just a prettier queue for the same slow line?
been there, migrated that