There are a lot of reference doc to tell about comparison between them
Reference:
https://cloud.google.com/architecture/building-production-ready-data-pipelines-using-dataflow-developing-and-testing#choose_correctly_between_side_inputs_or_cogroupbykey_for_joins
https://cloud.google.com/blog/products/data-analytics/guide-to-common-cloud-dataflow-use-case-patterns-part-2
https://stackoverflow.com/questions/58080383/sideinput-i-o-kills-performance

C -> Adding more nodes to a cluster (not replication) can improve the write performance
https://cloud.google.com/bigtable/docs/performance
D -> since Google recommends adding nodes when storage utilization is > 70%
https://cloud.google.com/bigtable/docs/modifying-instance#nodes

The key requirement is not to lose the data, the Dataflow pipeline can be stopped using the Drain option.
Drain options would cause Dataflow to stop any new processing, but would also allow the existing processing to complete.

Categorical features in linear models are typically translated into a sparse vector in which each possible value has a corresponding index or id. For example, if there are only three possible eye colors you can represent 'eye_color' as a length 3 vector: 'brown' would become [1, 0, 0], 'blue' would become [0, 1, 0] and 'green' would become [0, 0, 1]. These vectors are called "sparse" because they may be very long, with many zeros, when the set of possible values is very large (such as all English words).
[0, 0, 0, 1, 0, 0, 1] is not a sparse vector because it has two 1s in it. A sparse vector contains only a single 1.
[0, 5, 0, 0, 0, 0] is not a sparse vector because it has a 5 in it. Sparse vectors only contain 0s and
1s.
Reference:
https://www.tensorflow.org/tutorials/linear#feature_columns_and_transformations