Transactions have no intrinsic total order, which would be needed to have a chronological order. Two nodes might independently create transactions a and b and send them to a common nodes N and M. N receives first a and then b, while M receives first b and then a, which would you say is the correct order?

If there were a total order it would be possible to completely avoid the double spending problem as it would immediately be clear which is the winning transaction between two conflicting ones: just pick the first.

Going back to your original question: there are two separate sets of transactions you have to consider.

• Committed transactions: transactions that have appeared in a block in the longest blockchain
• Transactions in the memory pool: transactions that are yet to be confirmed

bitcoind provides access to both these sets:

• Use getmemorypool to get the transactions in the memory pool
• Use getinfo to get the height of the latest block, and then use getblockhash + getblock to iterate through the blocks in descending order and extract the transaction hashes. Stop when you have 10'000 transactions in total.