Why hyperdrives fail at 0.5 past lightspeed
The failure almost never happens at the speed you were worried about. It happens forty minutes later, and the cause was measurable before you left.
We have now torn down 214 failed motivators from customer hulls — ours and, more often, other people's. The pattern is consistent enough that I want to write it down plainly, because the folklore around hyperdrive failure is wrong in a way that costs crews money and occasionally costs them more than that.
The failure is not at the transition
Almost nobody's drive fails at the moment of transition to superluminal. Transition is violent, brief and extremely well instrumented; if something is going to break there, it breaks on the test stand. Of our 214 teardowns, eleven failed at transition. The other 203 failed between thirty and ninety minutes into a sustained run.
That timing is the clue. Thirty to ninety minutes is how long it takes a marginally mis-calibrated field lattice to walk itself out of true.
What actually goes wrong
Here is the sequence, every time:
- One lattice node sits slightly outside tolerance — often from a hard shutdown months earlier.
- At sustained 0.5, that node runs hotter than its neighbours.
- Hot node drifts further. Neighbours compensate, so the aggregate field reading stays green.
- Compensation exceeds the neighbours' authority. Aggregate reading drops in seconds.
- The drive drops you to realspace, hard, wherever you happen to be.
Step three is the dangerous one. Every conventional motivator reports aggregate field integrity, which is exactly the number designed not to move until it is too late to matter.
Read the node data, not the summary
If your drive publishes per-node telemetry, look at the spread between the strongest and weakest node, not the average. A spread above 3% is a scheduled maintenance item. A spread above 6% means you should not begin a long jump.
| Node spread | What it means | What to do |
|---|---|---|
| Under 2% | Healthy | Nothing |
| 2–3% | Early drift | Note it, watch the trend |
| 3–6% | Real | Schedule a re-trim |
| Over 6% | Pre-failure | Do not jump long |
If your drive does not publish per-node data, that is itself a finding.
The alluvial damper is not optional
The second most common cause in our teardowns — 47 of 214 — was a damper left in service well past its erosion limit. Dampers are consumables. They cost a fraction of a percent of the drive, they are the cheapest insurance in the propulsion bay, and crews skip them because nothing appears to be wrong. Nothing appears to be wrong right up until step four.
A hyperdrive that has never told you bad news is not a healthy hyperdrive. It is a hyperdrive that is not talking to you.
What we did about it
The HD-7 trims its weakest node against its strongest 400 times a second, which converts the failure mode above from a cliff into a slope. It also refuses to spin up on a badly seated damper, because the alternative — spinning up and finding out at minute forty — is worse.
None of that makes the drive immortal. It makes it honest, which in this business is the more useful property.