Ignore the tool involved as this can result from nearly any tool.
Some things you can try to “reduce” input volume/complexity for assembly tools:
Make sure you do some QA, then check the results of that QA, and do some more if needed. That usually means using FastQC twice (before and after).
Try using different assembly tools. Shovill with the SPAdes option has a very complete log report, even for errors, and often gives some clues about what you can do to mitigate the job running for longer or using more memory than is really necessary.
Sub/down sample the reads yourself instead of asking the assembler to do that, set the expected assembly size, and similar tunings will result in not just a successful assembly, but usually a better quality assembly result.
Search the GTN with keywords (quality, assembly) or navigate by analysis domain. Tool forms also have link-outs to the author publications/resources with advice. And, sometimes a search at general bioinformatics forums is worth it, too. Under the hood, the tool is the same as what everyone else is using outside of Galaxy, so job settings and data prep “best practices” that communities develop will still apply.
Extra note: Most assembly tools will expect intact pairs. Not all QA tools filter out the remaining mates when one end is rejected for quality reasons, or at least not with default settings. The tool Fastq info will check that for you.