9.40 reversion problem

Revert 9.4.TC6 to 9.21.TC5X6 on Windows 2000 Advanced Server SP4 Backend to a BaaN ERP system - as it's basically using it like a rather mad flatfile database - basically, none of the known issues with this reversion highlighted in TFM (nothing added since upgrade and not converted using -BCx) When we converted, the conversion process raised logs from 850 to 1360, but as I understand it, as it converts the logs anyway, this should not be an issue (other than the need to change the logs setting to reflect the change in our backup of the 9.21 installation). However, when we go into fast recovery, the system falls over, with the following entry in the online log; 15:08:39 Booting Language <c> from module <> 15:08:40 Loading Module <CNULL> 15:08:40 Booting Language <builtin> from module <> 15:08:40 Loading Module <BUILTINNULL> 15:08:45 Informix Dynamic Server 2000 Version 9.21.TC5X6 Software SerialNumber RDS#N000000 15:08:49 Informix Dynamic Server 2000 Initialized -- Shared MemoryInitialized. 15:08:49 Physical Recovery Started at Page(5:10341). 15:08:50 Physical Recovery Complete: 0 Pages Examined 0 Pages Restored. 15:08:50 Logical Recovery Started. 15:08:50 10 recovery worker threads will be started. 15:08:51 Logical Logging error for 'OLDRSAM:CKPOINT' in 'logread' 15:08:51 Assert Failed: Dynamic Server 2000 must abort 15:08:52 Informix Dynamic Server 2000 Version 9.21.TC5X6 15:08:52 Who: Session(8, informix@CHIMERA, 0, 0) Thread(17, fast_rec, 0, 3) File: rslog.c Line: 3203 15:08:52 Results: Dynamic Server 2000 must abort 15:08:53 Action: Reinitialize shared memory 15:08:53 stack trace for pid 1516 written to C:\\tmp\\af.3f9ed72 15:08:57 See Also: C:\\tmp\\af.3f9ed72 15:09:00 Releasing server from system block 15:09:24 PANIC: Attempting to bring system down 15:09:25 rslog.c, line 3203, thread 17, proc id 1516, Dynamic Server 2000must abort. don't know how relevant and useful this is, but AFAIK onmode -l and onmode -cweren't done before reversion (it's not stated in TFM that this is required, but personally I would probably have had a go first), anyhow, onstat -l output (snipped) from af. file is as follows; c4d98f4 555 U-B---- 439725 2b77531 1250 1250 100.00 c4d9910 556 U-B---- 439726 2b77a13 1250 1250 100.00 c4d992c 557 U-B---- 439727 2b77ef5 1250 1250 100.00 c4d9948 558 U-B---- 439728 2b783d7 1250 1250 100.00 c4d9964 559 U-B---- 439729 2b788b9 1250 1250 100.00 c4d9980 560 U-B---- 439730 2b78d9b 1250 1250 100.00 c4d999c 561 U-B---- 439731 2b7927d 1250 1250 100.00 c4d99b8 562 U---C-L 439732 2b7975f 1250 1044 83.52 c4d99d4 563 F------ 0 2b79c41 1250 0 0.00 c4d99f0 564 F------ 0 2b7a123 1250 0 0.00 c4d9a0c 565 F------ 0 2b7a605 1250 0 0.00 I know that there was (what looks like) a similar problem with earlier versions - but I understood this bug to be fixed several versions ago? Any ideas, suggestions?