PolymarketQuestion on GET /closed-positions (data-api). For some high-activity wallets the endpoint never signals end of pagination, so a client that stops on a short page paginates effectively forever (we hit repeated OOMs from this). Repro — user 0x204f72f35326db932158cba6adff0b9a1da95e14, limit=50: Full 50-row pages returned at every offset we tested, up to offset=1000000 offset=200000 and offset=500000 return the identical 50 rows, so offset seems to be ignored past some threshold (constant page from there on) offset=0 vs offset=200000 share 0 assets, so early pages do advance normally; the repeat only kicks in in the stuck region No wallet realistically has anywhere near that many closed positions Net effect: len(page) < limit never fires, and dedup-by-asset only partially helps (it catches the stuck region but not the large fresh-id range before it). A few questions: Is this a known regression? It seemed to start around late June / early July, roughly alongside the World Cup volume ramp. Is there a total/count field or a documented max offset we should rely on instead of paginating until a short page? Is there a keyset/cursor variant of /closed-positions (like /markets next_cursor) that terminates cleanly? Thanks!
Question on GET /closed-positions (data-api). For some high-activity wallets the endpoint never signals end of pagination, so a client that stops on a short page paginates effectively forever (we hit repeated OOMs from this). Repro — user 0x204f72f35326db932158cba6adff0b9a1da95e14, limit=50: Full 50-row pages returned at every offset we tested, up to offset=1000000 offset=200000 and offset=500000 return the identical 50 rows, so offset seems to be ignored past some threshold (constant page from there on) offset=0 vs offset=200000 share 0 assets, so early pages do advance normally; the repeat only kicks in in the stuck region No wallet realistically has anywhere near that many closed positions Net effect: len(page) < limit never fires, and dedup-by-asset only partially helps (it catches the stuck region but not the large fresh-id range before it). A few questions: Is this a known regression? It seemed to start around late June / early July, roughly alongside the World Cup volume ramp. Is there a total/count field or a documented max offset we should rely on instead of paginating until a short page? Is there a keyset/cursor variant of /closed-positions (like /markets next_cursor) that terminates cleanly? Thanks!