Thanks for the quick response.
On my initial prototype, I’ve used a 1% tolerance 330pF C0G capacitor for Ct. I don’t know the exact type (since I’ve had it lying around our lab for some years), but it is probably from Kemet. For Rfmin I’ve used a thin film 0.1% part from Panasonic, ERA3AED113V. For Rfax I’ve used a standard 1% Yageo thick film part.
According to the formula for fmin in the datasheet, these components should give a nominal fmin of about 92kHz. On my prototype the measured minimum switching frequency, measured with only the controller powered on, was 84.5kHz.
Today I’ve measured two additional proto boards, this time fitted with a known 1% 330pF C0G capacitor, an AVX 06035A331FAT2A. These board were again measured with only the controller powered on. These boards measured 84.2kHz and 84.8kHz. As a sanity check, I’ve removed Rss and Rfmax from one of the boards to completely isolate the fset pin, but this did not change the measured frequency.
It seems (from an admittedly extremely limited sample dataset) that the absolute tolerance of the oscillator is quite good, but maybe there is an error in the datasheet formula?