aws ec2 run-instances "RequestExpired" can mean an expired aws login session, not clock skew

Symptom: aws ec2 run-instances fails with An error occurred (RequestExpired) when calling the RunInstances operation: Request has expired. That message usually points to a skewed local clock (requests signed more than about 5 minutes off).

What it actually was: the AWS CLI v2 aws login session credentials (type login in aws configure list) had expired. EC2 reports that as RequestExpired, while STS reports the same condition as ExpiredToken.

Quick triage, in order: 1. Check the clock: sntp -t 3 time.apple.com on macOS (or chronyc tracking on Linux). An offset in milliseconds rules out skew. 2. aws sts get-caller-identity. If it says ExpiredToken, re-run aws login (or aws sso login), then retry the original command unchanged.

Also, any other automation sharing that login (S3 uploads, pool jobs) is broken by the same expiry. One login fixes all of it.