There were a series of fundamental constraints on the length of observations with Herschel. There was an operational constraint that the coolers on PACS and SPIRE had to be re-cycled for approximately 2 hours every 48 hours (a parallel cooler re-cycle took 2.88 hours, a SPIRE-only re-cycle took 2.52 hours and a PACS re-cycle 2.37 hours). However, in practice, the limit was imposed by the need to have a 3-hour daily telecommunications period (DTCP) with the ground station to download data and upload instructions every day and up to 3 hours each day were also reserved for routine calibrations. Thus there was a limit of 18 hours to individual observations with Herschel that was hard-wired into HSpot. Observers who wished to take longer observations than this had split their AOTs into shorter segments. Special care had to be taken when requesting observations close to the 18 hour limit that they would remain possible even after a slight change in the way that observations were executed on-board, or calibration strategy, as knowledge of the instruments improved in-flight. A significant number of observations did become unusable without modification during the mission and had to be revised to bring them back down under the 18 hour limit.

Very long AOTs -- particularly in SPIRE PACS Parallel Mode, for which there were few short observations that could be used as fillers in the telescope schedule -- imposed strong constraints on mission planning and could be difficult to accommodate in the telescope schedule, because they effectively filled an entire observing day and blocked it for other observations. It was often necessary to place either PACS or SPIRE phometry observations as fillers in the schedule, "wasting" helium that was being used to keep the unused instrument cool.

For PACS spectroscopy observations there was a significant danger that, for very long integrations, an instrumental glitch could cause the entire observation to be lost. Observers were advised to check with the PACS spectroscopy experts via Helpdesk if they were uncertain what to do to make thir observation more robust against a failure of this type. However, when very long PACS spectroscopy AORs were selected for scheduling, an automatic check was made by the Mission Planning Team, who would query the advisability of any very long integrations that appeared in the schedule with the relevant instrument expert. When an AOR exceeded 40000s duration, it was almost always broken into two AORs. As the instrument controller was re-set at the start of each AOR, even if it had glitched during the first half of an observation, the second half would still be executed correctly; this meant that only half of the observation needed to be re-scheduled, hence a significant amount of observing time was saved (and observations that would otherwise have been completely lost were saved) in this way.

For a photometric deep integration on a fixed target, the telescope could only stare at a single point in space for 50000s (13.9 hours) thus, the maximum point source photometry AOR length permitted by the software was significantly shorter than 18 hours. In practice, this was never an issue because, in the end, very little point source photometry was scheduled, with all point source photometry AORs being evolved early in the mission into mini-scanmaps.

Moving targets had to be dealt with in Mission Planning in a different way to fixed targets, as the spacecraft had to calculate an instantaneous position and track on it, rather than on the stars. This required the mission planning software to interpolate the position of the object at any moment from the Chebyshev Polynomials that define the target's ephemeris. This process may not be valid for integrations longer than 5 hours on fast-moving targets, hence the tracking accuracy cannot be guaranteed for longer moving target AORs; HSpot placed a blanket, hard-wired limit of 5 hours on the duration of a single AOR for solar system objects -- as usual, this limit could be avoided to splitting a very long observation into several shorter ones that were concatenated so, in practice, it was not a significant limitation for observers.