fix(infra): switch testnet DBs to x86

[xDarksome]

Description

ARM instances are totally unstable during data migrations. There's some issue in RocksDB on ARM making it periodically crash.

How Has This Been Tested?

Manually

Due Diligence

• Breaking change
• Requires a documentation update
• Requires a e2e/integration test update

[claude]

Claude finished @xDarksome's task in 1m 38s —— View job

---

Code Review

Issue 1: Commit Message Contradicts Code Change
File: infra/testnet/main.tf:135
Severity: Medium

The commit message "switch operator-b db to arm" contradicts the actual code change, which switches operator-b to x86 architecture. The change adds { cpu_arch = "x86" } to the db configuration.

Based on infra/modules/service/main.tf:59, when cpu_arch is not specified, it defaults to "arm". Therefore:

• Before this PR: operator-b used ARM (default)
• After this PR: operator-b uses x86 (explicit override)

The commit message should accurately describe switching FROM arm TO x86, not the reverse.

Issue 2: Inconsistent Pattern With Other Operators
File: infra/testnet/main.tf:110-149

After this change, testnet will have:

• wallet-connect: x86 (with clear justification comment at line 112-113)
• operator-a: ARM (default)
• operator-b: x86 (no comment explaining rationale)
• operator-c: ARM (default)
• operator-d: ARM (default)

The wallet-connect operator includes a comment explaining why it uses x86 ("so we have both architectures being actively tested"). However, operator-b's switch to x86 lacks similar documentation. Given the PR description mentions ARM instability during data migrations, this critical context should be documented inline for future maintainers.

Issue 3: Incomplete Migration Strategy
Context: PR description states "ARM instances are totally unstable during data migrations"

If ARM has fundamental instability issues with RocksDB during migrations (not just testing diversity), only switching 1 of 4 operators to x86 leaves 3 operators (operator-a, operator-c, operator-d) vulnerable to the same crashes. The PR description suggests this is a reliability fix, not a testing strategy.

Consider whether all testnet DB instances should migrate to x86, or clarify if this is a gradual rollout strategy.

---

[Copilot]

Pull request overview

This PR addresses RocksDB stability issues on ARM architecture by switching all testnet database instances to x86. Previously, only the wallet-connect operator used x86 for testing purposes, while other operators defaulted to ARM.

Changes:

• Sets default database architecture to x86 for all testnet operators
• Removes the special-case x86 override for the wallet-connect operator
• Minor formatting cleanup in IAM policy alignment

Reviewed changes

Copilot reviewed 1 out of 2 changed files in this pull request and generated no comments.

File	Description
infra/testnet/main.tf	Adds cpu_arch = "x86" to the shared db_config, ensuring all testnet database instances use x86 architecture; removes now-redundant override for wallet-connect operator
infra/modules/service/main.tf	Aligns formatting of Effect and Action fields in S3 IAM policy statement

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.