If two gzipped files are concatenated together, the result is a valid gzip file. However, it appears that pyarrow.csv.read_csv will only read the portion related to the first file.
If the repro script here is run, the output is:
$ python repro.py
pyarrow.csv only reads one row:
pandas reads two rows:
pyarrow version: 0.14.0