SQLite Forum

Multithread processing, storing in memory and writing to database
Login
> Time-series databases certainly do exists, but are there any as small and
beautiful as SQLite

Indeed, and its increasingly common to see SQLite at the endpoints in IOT etc for storing configurations and such, for ex, but then the already present SQLite cannot be used for time-series data and that data has to be stored in some other way for ex, in CSVs or specialized tools or data structures in application language and ETL-ed to a db with time-series features for aggregation etc.

While this does not address OP's immediate problem, I would deviate a little from the thread trying to find workarounds or alternative solutions for OP and suggest a new feature in SQLite instead.
PostgreSQL, which SQLite derives a lot of inspiration from, has a time series DB called TimeScale which uses two core concepts of hypertable and chunks - it uses BTrees underneath and keeps all relational features intact. It also claims very fast single-machine ingestion rates so it certainly seems possible with BTrees (I am no expert, though). 

Unless something in SQLite design rules it out completely, I think, the chunking approach might be a great feature to add to SQLite to improve SQLite ingestion rate and add basic timeseries capabilities. 

https://blog.timescale.com/blog/what-is-high-cardinality-how-do-time-series-databases-influxdb-timescaledb-compare/

https://blog.timescale.com/blog/time-series-data-why-and-how-to-use-a-relational-database-instead-of-nosql-d0cd6975e87c/

https://docs.timescale.com/timescaledb/latest/overview/core-concepts/hypertables-and-chunks/#hypertable-benefits