Document SQLAlchemy's Engine Kwargs' Behaviour

docs/integrations/databases/sqlalchemy.md

@@ -14,18 +14,37 @@ Install `logfire` with the `sqlalchemy` extra:
 
 Let's see a minimal example below. You can run it with `python main.py`:
 
-```py title="main.py"
-import logfire
-from sqlalchemy import create_engine
+=== "Instrument a Single Engine"
 
-logfire.configure()
+    ```py title="main.py"
+    import logfire
+    from sqlalchemy import create_engine
 
-engine = create_engine("sqlite:///:memory:")
-logfire.instrument_sqlalchemy(engine=engine)
-```
+    logfire.configure()
+
+    engine = create_engine("sqlite:///:memory:")
+    logfire.instrument_sqlalchemy(engine=engine)
+    ```
+
+=== "Instrument Multiple Engines"
+
+    ```py title="main.py"
+    import logfire
+    from sqlalchemy import create_engine
+
+    logfire.configure()
+
+    engine_one = create_engine("sqlite:///:memory:")
+    engine_two = create_engine("sqlite:///:memory:")
+    logfire.instrument_sqlalchemy(engines=[engine_one, engine_two])
+    ```
 
 The keyword arguments of `logfire.instrument_sqlalchemy()` are passed to the `SQLAlchemyInstrumentor().instrument()` method of the OpenTelemetry SQLAlchemy Instrumentation package, read more about it [here][opentelemetry-sqlalchemy].
 
+!!! warning
+    It's best to use the `engine` or `engines` arguments. If no engine is specified, then `instrument_sqlalchemy` may
+    only work if it's called before `sqlalchemy` is imported, in which case all engines are instrumented.
+
 !!! tip
     If you use [SQLModel][sqlmodel], you can use the same `SQLAlchemyInstrumentor` to instrument it.