The author guidelines for the Journal of Computational Physics do not specify a word limit for the abstract, saying only that it should be 'brief'. But should I impose a word limit myself? I have heard that some indexing services can truncate overly-long abstracts, though I've not come across evidence that this is this case.
An abstract should sell your paper, introduce the family of constructs/theies used and may indicate the key narrative/outcomes. It should end after that so that one needs to read the paper. The better journals usually ask for stricter limits. (Journals of ) Pure sciences probably have editors wanting to allow you to use complete sentences and sensible sentences even in the abstract so skipped the abstract word limit out of politeness..