Why does rio-tiler log Different ThreadPool worker counts?

[drnextgis]

I am using titiler-pgstac with MOSAIC_CONCURRENCY=20 and have enabled DEBUG logging for rio-tiler. When I request a tile, I notice something in the logs that I can't explain:

eoapi.raster  | [2024-08-23 11:47:18,124 +0000] [6] [DEBUG] rio-tiler: Running tasks in ThreadPool with max_workers=20
eoapi.raster  | [2024-08-23 11:47:18,129 +0000] [6] [DEBUG] rio-tiler: Running tasks in ThreadPool with max_workers=40
eoapi.raster  | [2024-08-23 11:47:18,135 +0000] [6] [DEBUG] rio-tiler: Running tasks in ThreadPool with max_workers=40

Are two different ThreadPools being used here? And why does one of them have twice the number of workers?

[drnextgis]

I believe I understand:

• MOSAIC_CONCURRENCY controls the process of obtaining data from a remote source.
• RIO_TILER_MAX_THREADS manages the concurrent reading of each raster, with a default value of 5 times the CPU count.

[vincentsarago]

Yeah there are two levels of thread workers

• mosaic items
• assets reading (because each items can have multiple assets)