Is exon order always preserved in splicing? Are there any cases in which the splicing machinery constructs an mRNA in which the exons are not in the 5' -> 3' genomic order? I'm interested any such cases, whether they involve constitutive or alternative splicing.

I don't have any literature to back this up but I doubt that it occurs (at least frequently).

For example, imagine a simple three exon gene. Upon splicing `exon 1` to `exon 3`, `exon 2` would be excised as part of the intron lariat and subsequently degraded. So in order for exon 2 to be spliced to exon three you would need to either have splicing between `exon 3` and `exon 2` in the lariat or another copy of the pre-mRNA. This is typically called trans-splicing but it only occurs in specialized systems such as spliced-leader sequences in _C. elegans_.