This indicates a cross-app internal networking failure or dropping packets. I used to use the shared CPU instance tier in the NodeJS app; later, I switched to performance-1x tier. Things improved, but the connection timed out issues appear occasionally. Things are so flaky; many days ago, my Redis pub/sub based multi machine tweak did not work out.
I checked your org to see if I could find your Redis app. I think I did, and its CPU usage has been heavily throttled for a while. This would certainly degrade your Redis’ performance. I’d try scaling your Machine up to performance-1x and see if that helps with the connection drops.
If not, we can explore other possibilities. I’d gently suggest adding more specific details about your problem and/or the troubleshooting you’ve already done. Logs from your Node.js app that corroborate the drops or test results showing how you determined it’s an internal networking failure would help staff and community members get on the same page.
The 50% mark on the CPU usage chart is easy to misread on a shared-cpu machine, because it isn’t the threshold where throttling kicks in. Shared-cpu instances work on a credit balance system: your machine gets a small guaranteed baseline (for your app specifically, that baseline works out to about 6.25% of a single vCPU, not 50%), and it earns “credits” while it’s idling below that baseline.
Moving your app to a performance-1x CPU (which isn’t credit-limited) should resolve this.