There is a particular kind of project meeting where everyone agrees the supply situation is difficult and then proceeds to plan as though it were not. The schedule assumes parts arrive when needed, and the contingency is a line labelled “expedite”.
When the upstream constraint is equipment rather than willingness, expediting does not work. Nobody can make a wafer appear ahead of the queue, and a premium paid for position mostly moves someone else back. What remains under your control is when you commit, what you commit to, and how much of the design depends on that commitment.
Committing earlier is the obvious move and the one most often skipped, because a forecast feels like an exposure. It is, and that is the point: a supplier allocates against the forecasts that came with a signature. Committing narrowly helps too. An order for a part in one package and one temperature grade is easier to fill than one spread across variants, and the volume variant of a part is almost always the one that ships.
The design-side version is to keep options open where it is cheap. A footprint that accepts two pin-compatible parts, a memory map with headroom so a density change is not a respin, a power stage that tolerates either of two controllers — each costs an hour during layout and buys a whole alternative supply route later. Build the schedule around the longest confirmed lead time in the bill of materials, not around the average, and treat any date that nobody has re-confirmed in a month as a guess.