Peter,
From this output, I don't see any evidence that the job failed to complete, or that the identified memory leaks are causing any errors in the assembly. Did you get output and does it look good?
The system requirements will vary extensively depending on the depth of coverage you attempt to assemble. In particular, we suggest that you don't try to assemble average coverage of more than about 250x using AV454. I believe there is now a downsampling routine included in the package, but it may be separate (I know we created one).
Unfortunately, the parent package of AV454, Arachne, is now minimally supported and the AV454 developer is no longer at Broad, so we are providing minimal support for this package going forward. I would suggest you try out our new assembler, Vicuna. It is primarily designed for Illumina data, but has a preprocessing module to allow it to handle 454 or unpaired Ion data as well. We find the performance on 454 data to be comparable to AV454, and the performance characteristics should be superior on less powerful hardware since, unlike Arachne, it was designed to be able to run on desktops.
Mike