Here is one way to think about it. Given a level of fitness, imagine two competing strategies:
(1) Aim for the highest constant speed that is physically possible, vs.
(2) Aim for the highest constant power output that is physically possible.

For a perfectly flat, windless time trial, strategies 1 and 2 coincide. With varying terrain, however, it turns out that the optimal strategy falls somewhere in between the two. During climbs, for example, your speed should be less than average, but your power output should be higher than average. This applies to sandpits, runups, etc, as well. So it's natural to be really huffing and puffing at the top of a run-up, but you shouldn't run yourself into oblivion (obviously). It's a matter of carefully doling out those matches.

http://sportsmedicine.adisonline.com/pt/re/spo/abstract.00007256-200737080-00001.htm;jsessionid=JG3QTLT1sjB8DwH37V4dQ1pNbsvb93ZzpBRvZ6MSDphDXXzgHKcB!-1891305337!181195628!8091!-1
When environmental conditions are relatively stable (e.g. in a velodrome) and the TT is >10 minutes, then an even distribution of work rate is optimal. For a shorter TT (<=10 minutes), work rate should be increased during the starting effort because this proportion of total race time is significant. For a very short TT (<=2 minutes), the starting effort should be maximal, since the time saved during the starting phase is predicted to outweigh any time lost during the final metres because of fatigue. A similar 'time saving' rationale underpins the advice that work rate should vary in parallel with any changes in gradient or wind speed during a road TT. Increasing work rate in headwind and uphill sections, and vice versa, decreases the variability in speed and, therefore, the total race time.