It's been brought to my attention that deep within the Panelbase datasets there seems to be the equivalent of a full-scale Scottish poll of Westminster voting intentions. It's presented as the subsample of a GB-wide poll, but it's derived from a proper-sized sample, so assuming it's been correctly weighted, the numbers can be regarded as credible. With Don't Knows excluded, they appear to be -
Liberal Democrats 5%
Nothing desperately surprising there, you might think (except for the Tories' unusually good showing), but what's interesting is that the traditional gap between Westminster and Holyrood voting intentions seems to have vanished in a puff of smoke. If this result turns out to be typical, Labour can't blame their low-20s showing in recent Scottish Parliament polls on "the Holyrood factor" - they seem to have genuinely lost even more support since last May. By the same token, anyone who has been thinking "the SNP got 50% for Westminster, so they're bound to do even better this year" is probably barking up the wrong tree.
The most important reason for the traditional gap was of course that some voters didn't think the SNP were relevant in a Westminster election. That no longer applies, so it's perfectly logical that the gap has gone.