A

422 Unprocessable Entity

usually means Prefect reached the API successfully, but the API rejected the contents of the request. So this is usually not a network outage. It’s more like: - a payload no longer matches what the API expects - a deployment/work pool config changed - a required field is now missing/invalid - parameters or job variables are failing validation Since you said everything was working until ~4pm Paris time, the most likely causes are: - something changed in Prefect Cloud / your workspace config around that time - a deployment or work pool was updated - a block / variable / credential reference used by your deployment became invalid - your code started sending a slightly different payload Most common Prefect cause: flow run creation or deployment execution with invalid

job_variables

/ parameters. What I’d check first: 1. Find the exact failing endpoint - Is it when: - creating a flow run? - updating a deployment? - setting state? - querying history? 2. Print the full response body, not just the exception headline. That usually contains the real validation error. If you can catch the exception, do this:

from prefect.exceptions import PrefectHTTPStatusError

try:
    # your Prefect call here
    ...
except PrefectHTTPStatusError as exc:
    print("status:", exc.response.status_code)
    try:
        print(exc.response.json())
    except Exception:
        print(exc.response.text)

That often reveals something like: - missing required field - invalid parameter type - schema validation failure - bad work pool job variable - invalid UUID/reference A few targeted things to inspect: - Deployment parameters changed? - Work pool base job template changed? - Referenced block deleted/renamed? - Environment / variables changed around 4pm? - Prefect version mismatch between client and server/cloud behavior? If you’re using Prefect Cloud, also check whether all prod processes fail, or only: - one deployment - one work pool - one worker type - one code path That helps isolate whether this is config-specific. I couldn’t confirm GitHub issues/discussions right now because GitHub access failed on my side, but based on Prefect 3.x behavior, this is typically a validation/config problem rather than a platform-wide “hard down” problem. If you want, send me: - the full traceback - the response JSON/text from the 422 - and ideally the exact Prefect call that triggers it and I can help pinpoint the likely root cause quickly.