Droplet randomly takes significantly longer than normal to send responses to requests.

My droplet (Basic / 2 GB / 2 vCPUs | Ubuntu 20.04.4 LTS) will ocationally reply to a web request incredibly slowly (static file). This happens on my normal nginx configuration with SSL, and in a test clean configuration of both nginx and Apache2 (no SSL), which I tested in clean lxd containers with the default config for both. This occurs in both Chrome and Firefox, the two browsers I checked it on. CPU, memory, and disk both do not appear to be maxed, with the highest CPU recorded being ~2%. My droplet is in region NY3.

Most requests work fine, but seemingly at random a request will take significantly longer for the entire duration of the request, not speeding up as other connections finish. In Chrome’s network panel, this time is listed as “Downloading Content”.

An example of request timing (from Chrome’s network panel):

This will ocationally effect large file downloads, turning a 5 minute download into an hour or longer. Cancelling the download and starting it again may fix the issue.

Is this to be expected with shared (basic) droplets? Is there anything that can mitigate this? I do not recall it always being an issue, but I do not know when it started.