--- title: Two people keep SQLite alive, Postgres quietly gets a... description: Today's batch is heavy on database internals and the operational scars that come with running them at scale. A profile of SQLite's maintainers reveals the... canonical: https://daily.dev/posts/two-people-keep-sqlite-alive-postgres-quietly-gets-a-plan-flipping-gotcha-z37td9hym twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Two people keep SQLite alive, Postgres quietly gets a plan-flipping gotcha | daily.dev og:description: Today's batch is heavy on database internals and the operational scars that come with running them at scale. A profile of SQLite's maintainers reveals the... og:url: https://daily.dev/posts/two-people-keep-sqlite-alive-postgres-quietly-gets-a-plan-flipping-gotcha-z37td9hym og:image: https://api.daily.dev/og/posts/Z37TD9hyM.png og:image:alt: Two people keep SQLite alive, Postgres quietly gets a plan-flipping gotcha og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Two people keep SQLite alive, Postgres quietly gets a plan-flipping gotcha **[Databases Digest](https://daily.dev/sources/databases_digest)** · 4 min read · 0 upvotes · 0 comments ## Summary Today's batch is heavy on database internals and the operational scars that come with running them at scale. A profile of SQLite's maintainers reveals the trillion-instance database runs on two full-time engineers and a 600:1 test-to-code ratio. A Postgres deep-dive explains why prepared statements can mysteriously slow down exactly on the sixth execution. Elsewhere, a Dolt engineer tears apart an ACM paper's benchmark claims, and ClickHouse details how it rebuilt its telemetry pipeline to survive an hour-long database outage without losing data. ## Content **TLDR:** Today's batch is heavy on database internals and the operational scars that come with running them at scale. A profile of SQLite's maintainers reveals the trillion-instance database runs on two full-time engineers and a 600:1 test-to-code ratio. A Postgres deep-dive explains why prepared statements can mysteriously slow down exactly on the sixth execution. Elsewhere, a Dolt engineer tears apart an ACM paper's benchmark claims, and ClickHouse details how it rebuilt its telemetry pipeline to survive an hour-long database outage without losing data. --- ## SQLite's trillion-instance install base is maintained by two engineers Richard Hipp built SQLite in 2000 for Navy destroyer software, and it's since become the most deployed database engine in history - more installs than MySQL and Postgres combined, by a wide margin. The numbers behind that reliability are wild: roughly 160,000 lines of C code backed by 90 million lines of test code, 100% MC/DC branch coverage, and DO-178B avionics certification. All of it maintained full-time by two people. [Read more](https://daily.dev/posts/yaPrnhLXL) ## Postgres prepared statements can silently flip plans on the sixth run PostgreSQL uses custom (parameter-aware) plans for the first five executions of a prepared statement, then on the sixth run compares the average custom plan cost against a generic plan and switches permanently if the generic one looks cheaper. That decision is based purely on estimated cost, never actual runtime, so skewed data can trigger a switch to a bad generic plan that sticks until invalidation. Drivers like psycopg 3 and JDBC auto-promote to named statements after 5 executions, which shifts the actual switch to around the tenth or eleventh call - worth knowing before you spend an afternoon chasing a query that 'randomly' got slow. [Read more](https://daily.dev/posts/A6UMIbUt5) ## A Dolt engineer debunks an ACM paper's prolly tree benchmark The paper claimed a 30-50% insertion speedup for DoltHub-style workloads, but the author found the comparison ran the paper's implementation entirely in-memory against Dolt's disk-persisted engine - not a fair fight. Two of the three proposed optimizations already exist in Dolt, and the third was never actually implemented or benchmarked in the paper's own code. Useful as a checklist if you've ever had to sanity-check a database paper before citing it in a design doc. [Read more](https://daily.dev/posts/FbYtmMDL6) ## ClickHouse rebuilt its OpenTelemetry pipeline to survive an hour-long outage LogHouse ingests 50 million events per second and 177 PiB of uncompressed data, and two earlier pipeline designs (agent-to-gateway, then local WAL files) both broke down under backpressure or operational overhead. The fix skipped Kafka entirely: an OTel failoverconnector routes to S3 only during backpressure, with a separate SQS-driven catch-up receiver draining the overflow once things recover. In a gameday test, the system survived a full hour of database downtime with zero data loss, while only paying object-store costs during the actual outage window. [Read more](https://daily.dev/posts/R8929yiPQ) --- ## Also notable - **Amazon Aurora PostgreSQL patches five version lines at once:** Aurora PostgreSQL-Compatible Edition now supports 18.4, 17.10, 16.14, 15.18, and 14.23, and AWS is explicitly flagging known CVEs as the reason to apply the minor upgrade during your next maintenance window. [Read more](https://daily.dev/posts/A3Wx1kzhv) - **Postgres log_temp_files defaults to off, and that's a work_mem blind spot:** A GUC deep-dive argues log_checkpoints and log_autovacuum_min_duration got sane defaults in Postgres 15 (on, and 10min respectively), but log_temp_files is still off by default despite being one of the clearest signals that work_mem is undersized. [Read more](https://daily.dev/posts/Ou8gWpgzH) - **Custom embedding models beat off-the-shelf APIs at under $4 a run:** If recall@10 on your domain queries drops below roughly 80%, fine-tuning a small open model like bge-base-en-v1.5 on a few thousand real query-document pairs typically beats OpenAI or Cohere embeddings by 10+ points, and the training run itself costs about $3.40 on an H100 GPU Droplet. [Read more](https://daily.dev/posts/0gOIp7lMx) - **Debezium's replication slots can grow disk until your primary dies:** A production CDC writeup on Postgres-to-Kafka-to-Iceberg pipelines flags replication slot disk growth as a failure mode that can take down a primary database, alongside concrete guidance on snapshot modes, exactly-once delivery since Kafka 3.3, and JMX metrics worth alerting on. [Read more](https://daily.dev/posts/43QqjUNc0) ## Similar posts on daily.dev - [SQLite for Everything](https://daily.dev/posts/sqlite-for-everything-rupb2mkye) · Lobsters · 7 upvotes · 1 comments --- Tags: [#postgresql](https://daily.dev/tags/postgresql), [#observability](https://daily.dev/tags/observability), [#sqlite](https://daily.dev/tags/sqlite), [#clickhouse](https://daily.dev/tags/clickhouse) [View this post on daily.dev](https://daily.dev/posts/two-people-keep-sqlite-alive-postgres-quietly-gets-a-plan-flipping-gotcha-z37td9hym) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"DiscussionForumPosting","mainEntityOfPage":"https://daily.dev/posts/two-people-keep-sqlite-alive-postgres-quietly-gets-a-plan-flipping-gotcha-z37td9hym","headline":"Two people keep SQLite alive, Postgres quietly gets a plan-flipping gotcha","text":"Today's batch is heavy on database internals and the operational scars that come with running them at scale. A profile of SQLite's maintainers reveals the trillion-instance database runs on two full-time engineers and a 600:1 test-to-code ratio. A Postgres deep-dive explains why prepared statements can mysteriously slow down exactly on the sixth execution. Elsewhere, a Dolt engineer tears apart an ACM paper's benchmark claims, and ClickHouse details how it rebuilt its telemetry pipeline to survive an hour-long database outage without losing data.","url":"https://daily.dev/posts/two-people-keep-sqlite-alive-postgres-quietly-gets-a-plan-flipping-gotcha-z37td9hym","datePublished":"2026-08-25T04:17:46.245Z","dateModified":"2026-09-13T21:34:12.052Z","author":{"@type":"Organization","name":"Databases Digest","logo":"https://media.daily.dev/image/upload/s--ir1hdV8b--/f_auto,q_auto/v1773839400/logos/databases_digest?_a=BAMAMiiu0","url":"https://daily.dev/sources/databases_digest"},"interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"isPartOf":{"@type":"WebPage","url":"https://daily.dev/sources/databases_digest","name":"Databases Digest"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Databases Digest","item":"https://daily.dev/sources/databases_digest"},{"@type":"ListItem","position":3,"name":"Two people keep SQLite alive, Postgres quietly gets a plan-flipping gotcha"}]} ```