worth checking your region first, quick question: is MonitorJogosWeb deployed in US West (California)? theres a confirmed railway incident there, "connectivity issues in us west," resolved 03:39 utc on aug 10 https://status.railway.com/incident/RL8FRJE6, right around when your latency started. their note says "a small number of services may need a redeploy to fully recover," and you only restarted postgres, not the web service itself, worth trying a full redeploy of MonitorJogosWeb if you are in that region.

if youre not in us west, this specific incident doesnt apply, and id look harder at your own connection pool theory instead, worth sharing your db pool size/timeout config

Thanks. I checked the deployment region and MonitorJogosWeb is deployed in US East (Virginia), so the US West incident does not apply to this service.

There were no code, configuration, or infrastructure changes on our side. The latency started suddenly from yesterday to today, without any changes to the application or database.

Do you have any other known or ongoing issue affecting web services or PostgreSQL connectivity in US East that could explain this sudden latency?

checked railways full status history for august, the only logged incident all month was that us west one, nothing for us east in this window. so no, i cant point you to a confirmed platform incident for this.

given that, your own connection pool theory is probably the better lead now. worth checking pg_stat_activity on postgres for any lock waits or long-running idle-in-transaction connections, and your app's pool size/timeout settings, if the pool got exhausted or a connection is stuck waiting on a lock, youd see exactly this pattern: normal cpu, normal memory, but consistent added latency per request. also worth checking if dns resolution to your postgres host is being cached or re-resolved on every connection, a slower dns path would add fixed latency without touching cpu

Thanks for the suggestions!

I tried both Restarting and Redeploying the MonitorJogosWeb service, but the issue persists.

Here are additional technical symptoms occurring right now in the Railway dashboard:

Exec Container Failure: When attempting to open the Shell console for MonitorJogosWeb, it fails with the error: WebSocket Disconnection: The Railway console repeatedly shows WebSocket connection failed when trying to establish a terminal session.

Latency: HTTP requests remain consistently slow (4-6 seconds), while CPU (0.001 vCPU) and Memory (0.042 GB) usage remain completely flat/low.

Since restarting/redeploying didn't clear the process block or restore access to the web shell, could this indicate an underlying host/node-level issue on US East, or a container process limit (NPROC/PID limit) lockup at the Railway runtime level?

Any insights on how to unblock the container shell session or check active connection threads directly would be greatly appreciated.

the redeploy resetting everything but not fixing it, plus the shell also failing, points away from your app and toward railways host/runtime for this specific deployment. worth escalating directly with these exact symptoms, thats not something fixable from the app side.

while you wait, add a temp endpoint that returns your db pool's connection counts so you can check state without needing the broken console

Hey, thanks for the detailed insight, that makes a lot of sense.

I just checked the database directly via DBeaver and can confirm the pool is not exhausted on the DB side. There are only 4 connections from my application, all in idle state with no locks or long-running queries. So as you pointed out, the issue is not on the app or database side, but on the Railway host/runtime for this specific deployment.

Given that the redeploy doesn't fix it and even the shell is failing, I'd like to escalate this directly to the Railway team with these exact symptoms. However, when I go to the "Get Help" option in the dashboard, the only choice I see is "Ask the Community". Is there another way to open a direct ticket or reach the infrastructure team? I want to make sure this reaches the right people since it's not something fixable from my side.

Thanks again for the help!

normally they do come inside comments too help like the mods and the employee i think you just have to wait one them will surely comment soon

Thanks!! I will wait for them. 😭