We analyse optical and X-ray spectroscopy of the Of?p star HD108, known for its strong dipolar magnetic field and its optical line profile variability with a timescale of $54 \pm 3$ yrs, interpreted as the stellar rotation period. Optical emission lines have now recovered from their minimum emission state reached in 2007 - 2008. The variations of the equivalent width of the H$\alpha$ emission provide constraints on the inclination of the rotation axis ($i$) and the obliquity of the magnetic axis ($\beta$). The best agreement between model and observations is found for ($i$, $\beta$) pairs with $i + \beta \simeq 85^{\circ}$ and $i \in [30^{\circ},55^{\circ}]$. The Balmer emission lines display stochastic variability at the $\sim 5$% level on timescales of a few days. TESS photometry unveils transient modulations on similar timescales in addition to prominent red noise variations. A Chandra X-ray observation of December 2021, when the star was at a higher emission level, indicates a slight increase of the flux and a spectral hardening compared to the August 2002 XMM-Newton observation, taken near minimum emission state. Magnetohydrodynamic simulations are used to compute synthetic X-ray spectra. With our current best estimate of the $\dot{M}_{B=0}$ mass-loss rate, the simulated X-ray luminosity and spectral energy distribution agree very well with the observations. Finally, the radial velocities vary on a period of 8.5 years with a peak-to-peak amplitude of 10 - 11 km s$^{-1}$, suggesting orbital motion with an unseen companion of at least 4 M$_{\odot}$.