[closed] Why would a linux server not support tcp SACK if the client does

background: I am troubleshooting a intermittent performance problem we have with a Windows 7 client connecting to oracle 10g database running on a linux 2.4.x server. [ note: i do not have access to the server ]. the app runs once a day as a scheduled task on windows 7. it makes numerous queries to the database and all the queries complete within 1 minute. the longest query is always the first one and that normally it takes 1-2 seconds.

this will happen for several days. then for a few days the first query will take 9.7 minutes to return the results, all other queries return quickly. then it will return to normal.

as part of that troubleshooting i noticed:

1. the following options were supported by the server during the handshake for the slow query: MSS, SACK, and WS (128)
2. the following options were supported by the server during the handshake for the fast query: MSS.

What would cause a server to support SACK for some connections and not support it at another time for the same client. In both cases the options supported by the client were the same.