@Charles Lewis good question — and yeah, you're right that the 1-second fix is a hot start scenario where the module still has valid ephemeris in memory.
For warm starts in that 10-15 minute window, the behavior really depends on whether the module retained its almanac and last known position. Almanac data stays valid for days, so even without fresh ephemeris it can use that to narrow down which satellites to look for and roughly where to expect them — that typically gets you a fix somewhere in the 15-30 second range rather than a full cold start crawl. Cold start from scratch with no cached data at all is where you'd see 45-90+ seconds depending on sky view conditions.
The inconsistency you're seeing might also be sky-view related rather than purely a cache issue — if your field sites have any tree lines or structures blocking elevation angles below ~15-20 degrees, you'll get variance in warm start times even when the almanac is intact, because the module might be hunting for satellites that are partially obstructed.
If you're hitting this regularly and it's affecting your mapping workflow, one thing worth trying is keeping the module in a low-power sleep mode rather than full power-off during short breaks — that way it retains ephemeris and you stay in hot start territory. Worth experimenting with on your next session.
