PostgresQueryBackend
A query backend of one aggregate type on PostgreSQL: the four primitives over one of the aggregate type's own tables, read as layout says, compiled to parameterized SQL (PostgresSnapshotSqlCompiler, PostgresAggregationCompiler): its snapshot table (on the bindings of PostgresSnapshotStorageAdapter; records of the shape of SnapshotRecord.toQueryRecord(), a row without any kept state field not reading the state at all) or its log table (on the bindings of PostgresEventStorageAdapter; one record per stream, as stored). A projection is applied to each record as it is read.
Every statement runs in its own read transaction under SET LOCAL statement_timeout of the admitted timeout; PostgreSQL cancelling it there is a StorageTimeoutException. JDBC calls run on dispatcher, virtual threads by default. A table not created yet (no snapshot of the aggregate type was saved, no stream appended) holds no records: every primitive answers as over an empty table, as the provisional facts of PostgresQuerySchemaAdapter describe it.
Functions
The records of query, read in full in their transaction, which then ends and gives its connection back, and only then emitted: a consumer as slow as an SSE client's never holds a connection of the pool the log appends through. The admission bounds the list (maxListSize); a budget without that bound (maxListSize = 0) holds a whole unbounded list in memory.