Before the FLCS timestamping and history changes, the reconciliation code zeroed the whole disk buffer for FLCS pages as a special case. For WT-8287 I left that alone, but marked it as something that could be improved later as it didn't seem to be particularly necessary.

It is, in fact, not necessary. There are only two pieces that need to be zeroed: (a) the last partial byte in the bitmap data, for bitmap sizes and page sizes such that the bitmap doesn't end on a byte boundary; and (b) the empty/waste space on an undersized page that includes timestamp data.