Just wondering whether or not Abyss a contig or scaffold is phased into a particular strand of the DNA (although synthetic).
My guess would be 'Yes' because
1. The tool uses pair-end information
2. When I align scaffolds to human reference, hg19 and check via IGV (http://software.broadinstitute.org/software/igv/), I see many regions covering by 2 scaffolds, each of which has different orientations (forward or reverse), similar to pair-end reads or 2 DNA strands.
No, ABySS does not attempt to create a phased assembly. Like most de novo genome assemblers, it is a haploid assembler (it collapses heterozygous regions into a consensus).
W.r.t. to your contig alignments in IGV, it is normal to see pileups in repeat regions, but you generally should not see coverage > 1 in non-repeat regions, unless you are allowing your aligner to produce multiple alignments per contig and are not filtering out poor-quality alignments.