I like the whole sqitch approach much better than that provided by most migrations (if I'm in migration land, I use https://flywaydb.org/). From the version control perspective you get the sprawl of changes across files like you do with migrations.

I find that, for tables, if I write anonymous "DO" blocks in a single script to manage both the initial creation of the table and any later deltas, I get a very satisfactory version control history. This is a bit different than the recommended approach (if I'm not mistaken).

I just wish it wasn't written in perl with 10,000 cpan dependencies which may or may not build in a given circumstance.