As you may know if you’ve ever used the UK Geoportal site to get boundaries as features or GeoJSON, once you’ve converted and re-projected your files you probably still have a file that’s too large to serve over the wire, even if you’ve simplified them with a tool like mapshaper.
A common approach to get around this is to split the file up into GeoJSON features, with a single feature per file, or load the whole lot into ElasticSearch. I might write a quick blog at some point about the Elasticsearch ETL process, but here’s a quick snippet to split out a file, in this case the Output Areas file, into thousands of GeoJSON files named by their GSS code.