Start by posting a copy of the [Invalid] result log. Most cases it gives a hint such as more frequent restarts or some other non-standard log line such as heartbeat issues.

We'll wait, and see ;>)

Useful comment ;o). 4 in 9 days are invalid, but no indication of the total completed in those 9 days to gauge the rate. Yes, invalids don't add to scientific value, but the effort is compensated for time in full. The result log will give us firmer indication and maybe the mention of the CPU... certain Mac CPUs are finicky in finding the matching wingman, meaning only when a result requires a wingman for any of the VINA sciences, but i5-2550 was not one of them FAICR.

Sorry for incomplete info; this may be only a little better.
HW: MacBook Pro from early 2008, the first model using Intel Core Duo (2.6 GHz) (Apple can't even give me a model # for this one). OS X 10.7.5, BOINC 7.0.31 BOINC is configd for only one CPU, so it processes a little less than 8 DSFL wus/day
Sitting on a laptop cooler and with thermal alarms to prevent frying.

Edit: I think I mistyped in my first post , it was more like 9 invalid in 9 days Not equally spaced.
The error rate by my calcs is therefore a bit more than 12% where it used to be 0%
The invalid logs are lengthy so I hesitate to post; more to the point I see no difference between them and logs for valid wus; starting up, lots of finished tasks and calling boinc_finish, no indications of warnings or restarts before that.
This refers to status logs on the results reported to the server; will dig into the files on the laptop and see if more info is there.

Edited to correct error-wu #

I don't think the length is a problem and could potentially lead to a solution. I haven't noticed any uptick on the Windows side of things. I have seen more errors by others on the work units that come my way. At times I am the third or fourth person getting a unit where others have gotten errors. It may be that some of the current units are more susceptible to minor changes while running.