Progress can be slow in repetitive regions

[dancooke]

Progress can be very slow in repetitive regions with mipmapped reads as many candidates are proposed. Often these regions are eventually skipped as there are too many possible haplotypes. The problem is especially bad for somatic and de novo calling where hours can be spent in a small region (< 1Kb). There is currently some logic in place to help avoid spending too much time in these regions, namely candidate generation masking, and disabling haplotype lagging in regions with too many candidates, but clearly these measures are not sufficient. We need a new approach to help detect these regions with high specificity.

[dancooke]

This is no longer such an issue in general, especially since 453c4ab. Somatic calling can still be slow in these types of regions (and quite slow in general), however this is more of a result of the tumour model in particular.