For those, who have large JIRAs (hundreds of thousands of issues) there are a few things to bear in mind when working with the Structure.
The recommended limit for the number of issues in one structure is 100K and with this Structure already might be working noticeably slower, especially if there are many users working with the Structure Board at the same time.
So what we recommend is to distribute the issues between several smaller structures (5-10K issues per structure works perfectly) - the number of structures does not affect the performance that much.
Another thing that may affect the performance are the synchronizers. Incorrect synchronizers configuration may lead to conflicts when one synchronizer reverses the actions of the other, and vice versa. There is a safeguard mechanism that stops the cycle and sends warnings, but the next user action might trigger it again, so there's a possibility of wasted CPU cycles and overgrowth of the structure change history.
There were also several customer specific issues, which only reproduced in the customer's environment, but they were successfully resolved each time.