Improve bounded jitter

[ThomWright]

Problem

Bounded jitter would jitter between min_retry_interval and the current interval. If the first retry always uses min_retry_interval then it would never jitter the first retry.

Solution

Use 50% of min_retry_interval as the lower bound.

Example:

• min_retry_interval = 1s
• First retry interval will always = 1s
• Current solution – will jitter between min interval (1s) and current interval (1s) resulting in no jitter
• New solution – will jitter between 50% of min interval (0.5s) and current interval (1s)

[tl-security-scanner]

Outdated security scan result

🔍 Security scanner results

[!WARNING]
A recent change introduced MEDIUM severity issue(s) in the PR, see the inline comment(s). Please review the issue(s).

Scanning is done on each push. If an issue is fixed, its inline comment will be removed.
To handle False Positives 🚫 or leave feedback 🐞 please reach out to Security Engineering on #ask-security-scanner - Thank you!
Scan id: [e90d5c86-d49d-4062-b36b-cb29c18e6fa3]

[jdon]

Bounded jitter would jitter between min_retry_interval and the current interval. If the first retry always uses min_retry_interval then it would never jitter the first retry.

Why is having the jittered_wait_for being potentially 0 causing issues?

[ThomWright]

Bounded jitter would jitter between min_retry_interval and the current interval. If the first retry always uses min_retry_interval then it would never jitter the first retry.

Why is having the jittered_wait_for being potentially 0 causing issues?

We might be misunderstanding each other.

The problem isn't that the jittered_wait_for is potentially zero – it's that it will never be zero (or anything other than min_interval.

Also, I just noticed a bug – will fix!