Feasible Pickup & Delivery job unassigned when time window is tight but valid

[shubham441996]

Description

The solver fails to assign a Pickup & Delivery job when the delivery window is tight, despite the solution being mathematically feasible and proven to work in runs with slightly broader windows.

Repro Steps

1. Run vrp-cli solve pragmatic pragmatic-unplanned.json --out-result unplanned-response.json --max-time 60 --heuristic dynamic --log.
2. Observe that Job 2168 is unassigned with TIME_WINDOW_CONSTRAINT.
3. Change the delivery window in unplanned.json from 10:40 to 11:15 (as seen in pragmatic-planned.json).
4. Run the solver again.
5. Observation: Job 2168 is now assigned and actually arrives at 10:08 AM, which is well within the original 10:40 AM deadline.

Expected Behaviour

The solver should find the 10:08 AM arrival solution for the 10:40 AM deadline, as it is mathematically feasible and requires no changes to the route other than shifting the depot departure time.

Root Cause Hypothesis

The solver appears to be defaulting to a later depot departure (09:22 AM in the failed run) and then fails to find an insertion for the Pickup & Delivery job because it doesn't "shift" the whole schedule earlier to accommodate the 10:15 AM delivery window. It also has 2 vehicles, which he can use to plan, but it is still unplanned.

Attached Files

pragmatic-unplanned.json
unplanned-response.json

planned-response.json
pragmatic-planned.json