In fairness, looking (from the outside) at the upgrade experiences of a number of #Mastodon nodes where I have accounts, the latest of which was an aborted 4.4 to 4.5 upgrade on one node (not this one) that lost a day's worth of data, I think that we can put this down to the general nature of the Mastodon upgrade process rather than the platform that it runs on.
Mind you, I have yet to hear of a large complex WWW-involved system with SQL-back ends where upgrades are *not* fraught with pitfalls. (-: