Adding late ASASSN-14ko peaks reduces support for a fixed Schwarzschild disc-crossing clock under constant delay

We test whether precession of a fixed Schwarzschild orbit can explain the flare times of ASASSN-14ko. The disc plane and mean delay from crossing to peak are fixed. Gaussian timing scatter is fitted. A clock with one emitting disc node fits 29 optical peak times through 2024 about as well as a cubic timing curve. Adding the January 2026 TESS peak changes its AICc score from 1.52-3.30 units below the cubic to 12.22-18.68 units above it. Higher scores are worse. The earlier fits predict this peak 17.28-19.30 days too late. These point predictions omit parameter uncertainty. Adding a separate SPHEREx peak without TESS gives a smaller score loss of 3.43-6.00 units, if infrared and optical peaks have zero offset. A clock in which both disc nodes emit fits poorly. Higher black-hole masses, a wider orbit search, and source light travel time do not rescue the full-record fit. Adding a three-day error floor retains the preference. Conditional simulations at the fitted one-node null give tails of 0.0050 and 0.0025 for two peak definitions that share observations. These are not uniform tests over all nuisance values. The package contains new orbit code, mass profiles, subset fits, simulations and numerical checks. This is a reanalysis, not new photometry. The result depends strongly on the latest TESS epoch. Moving discs, black-hole spin, changing delays and correlated noise remain possible. No orbital decay or common clock is established.