How long should one continue to transform from a single PCA fit?

Sorry if I'm asking that in a really odd or unintuitive way. Say I have data from a year ago and use the first months worth and extract the first 2 principal components for visual inspection on density-based clustering. I can use that same fit on the PCA instance to transform the data for the second month and third month and so on. But how can I determine if that change of basis is still appropriate (along the directions of …

clustering components data datascience extract visual visual inspection