WT-5406 helps us measure the impact of having double the number of transactions. We will come up with enhancements to optimize the code and reduce the overhead of having double the transactions.

From the discussions on the design ticket WT-5323, here are the possibilities we have thought of:

Considering the explosion of txn_global states:

• It might be interesting to consider if LAS read cursors could share a single WT_TXN structure.
• It might be interesting to consider if we could allocate LAS write cursor WT_TXN structures on demand, as many workloads won't need them.