Can you please clarify when you say we are not running normal stack..if it is because of my mention of running drupal 6.13, then apologies for the confusion..I shouldnt have mentioned Drupal..We are running Mercury 1.0 (Pressflow 6.17)

We first started with Mercury 1.1 on Lucid and as we started seeing the spikes on the server, decided to move the site to Karmic and Mercury 1

Here is the current server config

Ubuntu Karmic
Apache/2.2.12 (Ubuntu)
PHP Version 5.2.10-2ubuntu6.4
Mysql 5.0.83-0ubuntu3 (Ubuntu)
varnish-2.0.4
Mercury 1.0

But we still have the problem..
As suggested we did a Memory profile  and here are the results (http://groups.drupal.org/node/80274#comment-253904)

xdebug report

Showing the 20 most costly calls sorted by 'memory-own'.

                                                                        Inclusive        Own
function                                                        #calls  time     memory  time     memory
--------------------------------------------------------------------------------------------------------
MemcachePool->get                                                   88  0.2388 18694216  0.2388 18694216
module_list                                                        298  0.0718  9739040  0.0388  9304408
views_plugin_display->option_definition                            288  0.0374  9317408  0.0321  9295864
ob_start                                                           168  0.0040  7017848  0.0040  7017848
array_keys                                                        2200  0.0424  5640824  0.0424  5640824
func_get_args                                                     2284  0.0456  5168312  0.0456  5168312
str_replace                                                       5019  0.0886  5168008  0.0886  5168008
array_merge                                                        506  0.0130  4716592  0.0130  4716592
views_handler_field->option_definition                             250  0.0197  3983032  0.0154  3904872
drupal_load                                                        218  0.0624  4112664  0.0329  3890432
module_hook                                                      10694  0.5985  3513888  0.4189  3513888
t                                                                 2926  0.3052  3711784  0.1357  3484616
date_part_extract                                                  537  0.0274  3321384  0.0193  3291096
unserialize                                                        301  0.0192  2948712  0.0192  2948712
explode                                                           1449  0.0237  2667432  0.0237  2667432
variable_get                                                      4294  0.0896  2350280  0.0896  2350280
preg_replace                                                      3938  0.0864  1945864  0.0864  1945864
mysql_fetch_object                                                 915  0.0239  1840104  0.0239  1840104
_db_query_callback                                                3037  0.1914  1867120  0.1152  1757128
url                                                                651  5.0089  2557576  0.0876  1601952

Is there something else we can do to help isolate the issue?

Short CMS history of the site: The site had been running on a VPS on Drupal 6 from Jan of this year and before that on Drupal 5 for approx 2 years and on Drupal 4 for approx 2 years before that and was very stable. Approx 2 weeks ago, we moved to Rackspace cloud and pantheon and for the 3-4 hour period between load spikes, the performance of the site  is very very impressive. But when the load spikes, the site becomes completely unuseable and the only way to recover is to do a reboot of the server.