This work is dependant on support for text indexes on time-series collections.
We'd need to change the parsing behavior to specifically allow $text on a time series view. Currently $text must be with a $match at the beginning of the pipeline and is also specifically banned on views (although there is no custom error message).
We'd also need to ensure that the meta predicate pushdown code written for
SERVER-53484 handles text predicates. Tests can be added to pipeline/document_source_internal_unpack_bucket_test/
We should think about the ramifications of this patch. Currently optimization is entirely optional. How should we organize the code for a mandatory rewrite?