run_pgns may produce corrupt pgn files

[robertnurnberg]

The file 66fb690e86d5ee47d953b8eb/66fb690e86d5ee47d953b8eb.pgn.gz downloaded with the help of the run_pgns api contains the following corrupt sequence of games:

[Event "Batch 15: CM vs master"]
[Site "https://tests.stockfishchess.org/tests/view/66fb690e86d5ee47d953b8eb"]
[Date "2024.10.01"]
[Round "1"]
[White "New-7d27fed9884ce5204cf961b01ec5ea006ee83659"]
[Black "Base-d6043970bd156b1d2ab6cb51e8d5cb0c6a40797c"]
[Result "1-0"]
[FEN "r1bqkb1r/p1p1pppp/1pn2n2/8/2p1PP2/2N2N2/PP1P2PP/R1BQKB1R w KQkq - 0 6"]
[GameDuration "00:02:36"]
[GameEndTime "2024-10-01T01:46:13.286 -03"]
[GameStartTime "2024-10-01T01:43:36.979 -03"]
[PlyCount "118"]
[SetUp "1"]
[Termination "adjudication"]
[TimeControl "51.376+0.514"]

1. Bxc4 {+1.00/25 5.2s} Na5 {-1.37/24 6.9s} 2. Be2 {+1.18/26 3.2s}
e6 {-1.21/21 0.62s} 3. d4 {+1.31/21 1.3s} Be7 {-1.20/21 1.1s}
4. b4 {+1.31/20 0.87s} Nc6 {-1.30/21 1.6s} 5. a3 {+1.44/19 1.2s}
O-O {-1.25/23 1.4s} 6. Bb2 {+1.36/21 1.8s} Nb8 {-1.23/26 6.0s}
7. Qc2 {+1.40/22 1.1s} Ng4 {-1.55/32 12s} 8. h4 {+1.31/23 4.7s}
c5 {-1.49/20 0.51s} 9. e5 {+1.52/22 0.82s} Ne3 {-1.39/28 3.8s}
10. Qe[Event "Batch 149: CM vs master"]
[Site "https://tests.stockfishchess.org/tests/view/66fb690e86d5ee47d953b8eb"]
[Date "2024.10.01"]

As can be seen from https://tests.stockfishchess.org/tests/view/66fb690e86d5ee47d953b8eb, the batches 15 and 149 originate from two different workers. It seems that worker 15 uploaded an incomplete pgn file, which is then merged by the api with all the other batch pgn files.

A possible way to fix this would be to reject pgn files from workers that terminated in an unnatural way.