{Edit: One day later... still no sign of it. I suggest raising an issue with support if you need this fix.}
Apparently IBM has put a hot fix package on the Boulder FTP site. (I'm not sure how they'd feel about me putting the actual link up here, so I shall refrain from doing so for the moment; if you've downloaded hotfixes before you'll know where it is.)
I say "apparently" because the site appears to be down at the moment and I can't confirm this. Apparently (again) it's in the im/mike folder.
In any case, the hot fix addresses the following issues. The Undo/redo one would appear to be the critical one:
TM1 9.5 Hot Site (Dec 1, 2009)
With the release of TM1 9.5 on December 1, 2009, we are also shipping a pre-emptive hot site to address 5 known issues with the eGA release. Many of these issues will not be apparent in development or test environments and all of these issues are TM1 server related (not TM1 web):
Named Levels with BI [CQ 224252]
When using TM1 with C8BI with named levels, BI reports will not work if the named level for a dimension happens to match a member name.
Workaround: Rename the Named Level for the Dimension to ensure that it does/will not match any member of that dimension
Undo/Redo [CQ 221118]
If a "savedataall" is performed, it will adjust the log file in a way that essential disables the capability of undo/redo.
Workaround: The log file generated by the "savedataall" must be deleted in order for Undo/Redo to work properly again.
DB2 Relational Reports [CQ 225804]
If (1) running BI relation reports using DB2 first and (2) then a TM1 report in BI, the reports for TM1 will never work. This relates to how DB2 and TM1 instantiates the OpenSSL dll.
Workaround: Run the TM1 reports first and everything will work fine for that BI server session.
AIX [CQ 225341]
When running a AIX server under heavy load for several days, the TM1 server may crash. There are no exact reproduction steps for this crash other running the server for several days under AIX.
Workaround: None. This will most likely never be encountered in a development or test environment. In fact even in production, the customer may never see the effect of this bug.
Performance [CQ 218361]
Under heavy concurrency of over 100 concurrent users, TM1 9.5 will operate more slowly than TM1 9.4. This becomes more apparent as there are more concurrent users on a TM1 server. It appears that views with Dynamic Subsets causes most of the (but not all) of the performance impact.
Workaround: Performance can be improved to a degree by NOT using dynamic subsets if possible.
