After reading Ray's reminder that the old kep bulletins are online, I tried an experiment running SatPC32 WinAOS for a 24 hour period using the current keps (from last Friday's bulletin) and a set from the end of March (~45 days, which is about how many days there are until Field Day give or take). Summary of results for all the regularly operating FM and SSB satellites I can reach here and ISS:
78 passes at 5 deg or better in 24 hours Excluding ISS, largest AOS difference was 1 minute (assuming SatPC32 rounds to nearest minute, call it 1:30). Differences in predicted period (length) of pass were 1 minute, including ISS; roughly evenly distributed between shorter and longer. Differences in predicted max elevation was up to 7 degrees, average was 0.95 deg. The largest differences were associated with higher elevation passes, so not a big impact on ability to use.
I didn't check for horizon skimmers where a small change in the calculations could result in no pass or a marginal pass.
Conclusion: For general planning purposes, keplerian element data is useful at least 1.5 months out. Possibly longer. I'd be much more careful for planning a max distance attempt (horizon skimmers at both ends) and old keps are probably useless for computer-control of radios and antennas.
And my planning for Field Day is going to be relatively unchanged.
73 Steve KS1G