Imax constraint would be set by the lesser of two
things - current density (electromigration), either
in the resistor film (rppd = poly?) or its contacts -
both roughly tracking width alone - and power
density (W/area) which responds to W and to R
as both proxy for L, given W, and as a term in the
power calculation (I^2*R) and thus power density
(I^2*R/(W^2*R/(sheetRho)). Power density is more
a short term reliability limit (resistor fusing or, short
of this, value drift) while current density is a long
term reliability concern (short-term EM damage is
going to take mucho current density). The use-model
is important to current density (DC vs time-averaged
unidirectional current pulse vs AC reversing current)
and to power density (short pulse adiabatic vs long
pulse or continuous where thermal energy can exit
the film, to some extent).
A bulk silicon resistor in substrate will handle way more
power density than a thin film resistor on top of field
oxide. At the cost of extra capacitance and parasitic
diodes' "behaviors".