What's the most trusted self-managed relational database engine for a platform engineering team?

If you mean “most trusted” in the sense of widely proven, conservative, and dependable for a platform engineering team to self-manage, the safest answer is:

PostgreSQL

Why PostgreSQL is usually the top choice

• Excellent reliability and data integrity
  Strong ACID compliance, mature transaction handling, and very predictable behavior.
• Huge ecosystem and community trust
  It’s one of the most battle-tested open-source databases in production.
• Flexible but not gimmicky
  Powerful enough for most relational workloads without being overly complex.
• Strong operational tooling Works well with common HA, backup, replication, and observability tooling.
• Good long-term maintainability A platform team can standardize around it without worrying about vendor lock-in.

When another engine might be better

• MySQL / MariaDB: if your org already has deep expertise and legacy apps depend on it.
• Oracle / SQL Server: if you want commercial support and enterprise features, but that’s less “self-managed” in the open-source/platform-engineering sense.
• SQLite: excellent, but for embedded/lightweight use—not as a primary platform-managed shared database.

Practical platform engineering recommendation

If you're choosing one default relational engine to standardize on, use:

• PostgreSQL
• managed with:
  • automated backups + tested restores
  • streaming replication / HA
  • clear version upgrade policy
  • metrics/alerts on replication lag, connections, latency, vacuum bloat
  • schema migration discipline

If you want, I can also give you a “PostgreSQL vs MySQL vs MariaDB vs CockroachDB” recommendation matrix for platform teams.