Check your `compress_segmentby` choice. The reporter's root cause: the segment-by date column had nearly unique values, so compression could not group rows and the compressed chunks ended up larger than the originals. Pick a low-cardinality column for `compress_segmentby` (something with few distinct values, like a device or region id) and re-test on a chunk. If the old badly-compressed chunks are no longer needed, drop them with `drop_chunks()` before decompressing and reconfiguring, so you do not keep paying for the inflated storage.
_Source: gh:timescale/timescaledb#5961 (Timescale Cloud)_