Context Navigation

Changes between Initial Version and Version 1 of Ticket #36643, comment 7

initial	v1
3	3	I believe things are this way because the `migrate` command, with and without `--fake`, makes graph manipulation expecting the history graph to be in a consistent state prior to making these changes and that the operations it performs could venture into undefined behaviour territory without these per-requisite.
4	4
5		To make a ''typing'' analogy, if a piece of code expect a graph to follow specific schema (e.g. single head per app, acyclic) and you force it through the pipe in an unexpected form then it can ~~result~~ in damageable ways.
	5	To make a ''typing'' analogy, if a piece of code expect a graph to follow specific schema (e.g. single head per app, acyclic) and you force it through the pipe in an unexpected form then it can behave in damageable ways.
6	6
7	7	Given undefined behaviour is not something we want anywhere close to database schema alterations I believe the current approach is reasonable.