Sorry, I should have given a more detailed comment. I don't know of a specific tools release that guarantees that reconnect from a running JAS to a restarted enterprise server will work. I have a site on 8.96F1 and we find that usually we do not have to restart the JAS after an enterprise server restart. However as a matter of practice, in production we usually restart everything in sequence because we have found that the failover/retry behaviour of E1 is inconsistent.
Take a look at this white paper to get some official results from Oracle internal testing on failover:
http://www.peoplesoft.com/corp/en/doc_archive/red_paper/rp_e1perf_rac.jsp
The document is specific to Oracle RAC (Real Application Clusters) but the observations apply to pretty much any failover scenario with E1. I saw an early version of the document and the basic conclusion I made at that time is that transparent failover is not guaranteed.
Depending on what a particular user is doing at the time they are "reconnected" to an enterprise server there are four basic scenarios:
1. The user will not notice the reconnect and the active transaction will complete successfully while writing valid/complete data to the transaction tables.
2. The user will not notice the reconnect and the active transaction will "appear" to complete successfully while writing *invalid/incomplete* data to the transaction tables.
3. The user will be presented with an error message. Exiting the specific application and restarting it (not the E1 session) will allow them to continue working.
4. The user will be presented with an error message and will have to logout and login to get back to a valid session state.
The second experience is the most dangerous as the user is not aware of any problem and there may be data corruption after a failover.
Based on Oracle's own findings we should definitely keep restarting the JAS after a service restart or node failover (at least in production):
"EnterpriseOne interactive clients and batch processing UBE’s produced various results that were not consistent across the EnterpriseOne applications tested when the RAC node fail-over occurred. If Oracle 10g R2-RAC is implemented, users must be aware of the normal EnterpriseOne application interactive behavior and be very observant to visual differences that might indicate a node failure has occurred.
When the Oracle database RAC option is implemented a data verification process needs to be previously defined, tested, and accepted as a reliable method to be able to check the integrity of the data that was being accessed when the RAC node fail-over occurred. The methods implemented to perform this data integrity check may be an automated job, a manual interactive user process or a combination of both that every company implementing RAC must decide on how to best perform this check."
Again, although the above is Oracle RAC specific the findings definitely apply to the other supported E1 platforms/DB's.