While heights do change slowly by natural processes such changes are likely to be on the order of centimeters barring any extreme events.
Most changes in published heights of hills since people started measuring these things accurately (only the last couple of hundred years) are due to improvements/investments in technology.
Measuring the height of a hill is surprisingly hard. One of the earliest ways to determine hill height was trigonometry. If you have two points with a known distance between them and one known height (sea-level works well here) then by careful measurement of angles you can determine the height (and distance) of a third point. The accuracy here will depend on the the ability to measure an accurate angle and the distance to the third point. I haven't done any accurate calculation but with tools of the late 19th/early 20th c. I reckon accuracy of a few meters is reasonable. Possibly a bit higher with careful repeated measurements.
Interestingly this is what lead to the introduction of the system of trig points in the UK and other places. Which used this principle to map out distances and heights across the country.
Even in modern times calculating heights is not trivial. Here is an interesting article but the ordnance survey describing the process.
Primarily, they use photogrammetry (and probably related techniques such as LIDAR) to measure height by air. This is a quick way to cover large areas by is limited to accuracy of a couple of meters.
If you want to measure height accurately you need to leave a fixed GPS receiver in place for several weeks. While the accuracy here can be a few centimeters it is too expensive to be used in most situations.
Therefore there is significant uncertainty about some Munros height considering how close they are to the 3000' (914m) cut-off. The two post 1997 Munro re-classifications have both been due to interested parties hiring an team of surveyors to do very accurate measurements of some mountains which are more accurate than what the OS would normally do and they found that these two are just below the cut-off.